In case you are among those millions who had something important planned last friday, and were affected by this system outage, I'm sorry for you 😢

The highlights of the week:

BSOD Friday​
​In a nutshell: Crowdstrike launched pushed an update that crashed every windows system that was using it. Causing major outage across the world, and the fix isn't easy.

Crowdstrike - the first to knee half of the world

Disclaimer: First and foremost, I should mention this isn't directly related to AI as far as I'm aware. However, it's absolutely vital for every AI system out there because Operating Systems are the foundation upon which everything else is built.

What happened?

This friday 19th of July 2024. Some chap at this company Crowdstrike released an update that crashed the windows systems of every customer they have.

Wait, who is Crowdstrike again?

This company rolls out antivirus software that gets up close and personal with your system's core—think Norton or Avast. It's built to keep malware at bay on Windows-run machines. Here's the kicker: they're big on catering to the heavyweight enterprises, so their software is pretty much everywhere. Picture this: they're on board with around half of the top 1000 companies in the US alone, plus they've got a massive following globally. No surprise, when they hit a snag, it's not just a local hiccup—it's a worldwide event. 🗺️

How are they crashing windows?

Basically, someone goofed up and coded something that's reaching for a chunk of computer memory that's just not there. So what happens? The program throws a fit, Windows can't handle the drama, and boom— you're stuck in a reboot loop that just won't quit. 💥

The fix is easy but not really scalable.

So here's the deal: the fix is pretty simple. You just gotta fire up Windows, dig into the CrowdStrike files, and zap one of them. But here's the twist—you need to boot up in safe mode, which means someone's gotta be there to do the dance manually. Now, imagine this fix times a million, or heck, even a billion computers. That's a lot of folks who have to roll up their sleeves. 😳

If Bitlocker's in the picture, throwing passwords around like confetti, well, that's a whole different ballgame, because every machine's got its own secret handshake.

Who was affected?

Airlines, Banks, Stock Exchanges, Hospitals, Transport Systems etc. 😢

Conclusion

Last Friday was a real eye-opener—you know how a deck of cards can tumble with just a bit of a breeze? Turns out, the world's infrastructure is pretty much the same. A tiny nudge here or there, and bam, everything comes crashing down.

Hollywood's all about those flashy doomsday scenarios—aliens, mega-storms, scorching heatwaves, or a new ice age. But the real plot twist? It's us. Yep, we humans could send the world into a tailspin faster than you can say 'Mach 1'.

We really live in the future, one wrong button pressed, one person empowered to do something he shouldn't and we're done. 💣

Now to raise the mood a bit, because this is a little too dark today here are some memes:

That's it folks, see you next week ❤️️️️️️️

Happy Prompting!

​

Remember to invite your friends to subscribe at https://​newsletter.texti.app​

Europe's crying and celebrating at the same time, because Euro 2024 is done.

The highlights of the week:

1. Euro 2024 finals​
   🇪🇸 Spain won 🏆 the championship vs England 🏴󠁧󠁢󠁥󠁮󠁧󠁿. Fun fact England stole the flag of Genoa (a province in Italy) because they were the strongest maritime force back in the day, and everybody was afraid of them, and in order to protect themselves, Brits put Genoa's flag on their ships to avoid pirates in the sea. The more you know 😅
2. ​Kling​
   A catchy name for a rising Video Generation tool!

​Kling - the hot kid on the block

I wasn't really considering to write a post about Kling really, since it's yet another video generation model. You write a prompt and apply some styles and voilá you have a video. I mean we've already seen examples of this technology so many times this year. We've seen an attempt from google in Veo, Eleven Labs, Tavus.io, etc. All of them aim to build the ultimate video generator. However I think I found my favorite attempt so far.

I've seen this video on Twitter and well frankly it impressed me. I think this is one of the best takes on video generation in terms of day by day use of general public. In this video we can see how people used 2 pictures across generations and the video is generating a sentimental and sweet hug between the 2. I mean don't get me wrong it won't satisfy you physically by any means, however it would offer your the chance to see yourself next to your relative across time. It is a fantasy from Sci-Fi movies, we've seen, where a more immersive approach on this in many futuristic movies was shown, yet these examples make me think that an experience like this, is within our life-time.

One more example I found and I find it pretty fascinating is the following example. I'm honestly impressed that the objects in the image are interacting with each other. I remember when I was trying the first attempts on image generation, each subject in the image lived it's own life, now though, these are actually interacting. I really am impressed and I want to see more of it in the future.

My moto is: we live in the future, and I really mean it. Time passes, as you read this newsletter, you've already seen things that you maybe didn't know were possible, but also who you were 1 minute ago, is not the same person as you are now. We constantly change, and it's our decision to evolve or degrade with time.

Outro

A little off topic. Kling is a Chinese company, and things work differently there, to access it you'll need a specific process and a couple more strangely necessary apps. But worry not I found a detailed guide on how to access it in case you want to do it.

Tyyle @tyyleai Cats out of the bag, I know how to get access to Kling without having my own Chinese number!😛 A step by step thread on how to access Kling!🧵👇 8:47 PM • Jun 17, 2024 8 Retweets 39 Likes Read 4 replies

​

That's it folks, see you next week ❤️️️️️️️

Happy Prompting!

​

Remember to invite your friends to subscribe at https://​newsletter.texti.app​

First of all I have been working on this for a while, but from now on you will be able to receive the news in your favorite language. Please bear with me. I will still write the news in english and leverage the power of AI to translate them into your preferred language, it should still maintain the same character and identity. However, in case you see something wrong please let me know.

If you want to change the language you receive the newsletter please reply to this email with your language of choice.

Let's jump into today's news!

Figma Introduces it's own copilot

What did Figma just do? Well it started somewhat of a revolution for designers.

How did Figma out of the blue added a whole set of new AI features into Figma without rolling them one by one first? They're not Apple they can't do it.

Figma isn't Apple but they have the budget to copy something that Apple does 😅. They simply bought a startup called Diagram, that was working on AI features, and made it part of Figma, and voila a bunch of new features have launched 🪄.

Content Summarization

I will use this feature pretty often to be honest. In a parallel universe where I'm a clerk at a company, and make money to get rid of my hunger, I write a bunch of sticky notes in figma, most often weekly, in different meetings. Summarizing them like this is extremely efficient and really helps extracting the core message out of them.

A little about me: I move a lot, and I have been living in multiple countries, and have been experiencing all sorts of people and cultures. At a moment in my life, I have made a large chart adding pros and cons for different countries to live in (subjectively of course), and summarizing those in a manner of this sort with AI would've been fantastic, sadly I had to make a decision manually like a human...

My message is use this shit as much as you can, exploit it for your own productivity.

Variant Generation

It is not a secret that first and foremost Figma is a design instrument and designers use it at a large scale to generate consistent and effective products all around the world.

However you might not know that designers also have routine things to do from time to time, it's not all 100% creativity. For example if you have to move from blue buttons to green buttons, then you have to go and change all the designs manually. That was eventually fixed by a design system, change in one place to change everywhere. Duh – smart!

Another thing designers have is being hit by a creativity crisis so they usually research and check out what does the world do.

Figma is solving this problem as well, with a new cool tool, this allow you to generate variations of designs for specific areas of your product, see the gif above.

I do not know how consistent with your design these are going to be, or how wonderful these designs will be, but it might be more often then others that the instrument might provide an inspiration and help you out move from a dead spot.

For those who think that designers won't be needed, I'm sorry to inform you that this isn't the case. A tool to solve a problem will be easily adopted by people, while a replacement for a human, will likely start a revolution.

Currently Figma is providing a 10x productivity tool, use it and become better at what you do, because the future is bright and there's place for anybody who desires to participate in it. If you're reading this newsletter, you want to be a part of it.

Be bright ✨, be smart 🤓, be the future 🚀

That's it folks, see you next week ❤️️️️️️️

Happy Prompting!

​

Remember to invite your friends to subscribe at https://​newsletter.texti.app​

This week the AI hype is thrown all over the place, from windows to Amazon's Anthropic stuff.

The highlights of the week:

1. Window's new spy feature: Recall
   First of all this is literally an implementation of an AI Spyware. It screenshots your screen every-time it tracks an activity, and it's contents are kept in plain text 🤦🏻‍♂️
2. Claude 3.5 Sonet by Amazon​
   Another day, another model, now the bar is to beat GPT-4o, and it claims to do so.

​Windows Recall​

They better recall it. Duh, while this feature could be potentially a wonderful thing for humanity, it's execution is as poor as the name. What do you call a recall? Holy cow, no company ever thought of a recall in a good way. One of the worst marketing term one could've invented 😅

Anyways what is this, and what's the fuss about it?

1. It takes screenshots of your screen, every time it sees an activity;
2. It recognizes the text, and saves it in plain text;
3. It then allows you to recall what have you done 5 minutes ago by searching through-out your search history.

I can't see anything flawed in this work-flow 😂... One of the most critical things here is, it will definitely take screenshots of your credit card details and keep them in plain text.

Imagine somebody being able to just look at your entire history of what you did at your computer literally step by step, this is exactly this. In case you ever loose your laptop, loose the access to your computer, you're simply put - doomed 😈. As the attacker will be able to just check-out all of your history, and this person can impersonate you, better than you could, because all he needs to do is look into your history. Reminds me of that movie with Sandra Bullock where she got her life hacked.

Damn, can't even think of a worse feature to be implemented. We all know how secure is microsoft and it's reputation for disallowing intruders into your computers.

Wait, it's even worse, if you use a company computer, the System Administrator, and upper management, will always be able to just check whatever you did remotely, even without your consent, because that's how company computers work. So if you slacked or said something bad about your boss in a company chat, now everybody will know about this! Now I can't talk shit about my boss 😱!

Lastly, something that is kind of dumb and makes me laugh is the system requirements for this feature:

This means it'll make your computer slow, simply because it has to process all that information in the background.

What a move microsoft 😂

​Claude 3.5 Sonnet​

What is this and how can you eat it? It's a new model that can execute mathematical functions better and faster than we could previously do. Translating into better accuracy and faster response time. I mean OpenAI had fun during their presentation, presenting GPT-4o translating from one language to another. It made it looks actually good. Anthropic wrote a blog post. I guess everybody treats fun differently. Says the person who is writing a newsletter for fun... why would I even complain 😂

Anyways text generation aside they presented a really cool feature called Artifacts. What this means, is that Claude can now generate files you can run on your computer. Their demo is pretty sick. They generate a video game from a couple of prompts, and frankly looks really neat. The game should be called Mr. Krabs hunting shells without an internet connection.

​

It's impressive that the algorithm is able to write the code without any errors and the user could run easily, that said the demo looks heavily edited and could hide behind a lot of manual labor 🤷

Anthropic can make an Google like ooopsie and hide all the heavy lifting for investor reasons, we'll find out soon enough, when people start complaining about it on twitter 😅

That said the idea of Artifacts as is, sounds very cool. We'll definitely see this persistent all over the place, because it transfers the Large Language Model, from a text generation into something way more impressive and capable, it is now truly a comprehensive assistant for users.

I could even say that in a very near future OpenAI will re-label this feature and sell it to the majority of users.

That's it folks, see you next week ❤️️️️️️️

Happy Prompting!

​

Remember to invite your friends to subscribe at https://​newsletter.texti.app​

First of all, I'm extremely sorry, for my absence the past couple of weeks. I got lost in translations of AI. It took me so much time to figure out how to make this happen that one could think I dropped doing my thing. Don't worry though I'm not that easy to beat!

The most important:

Apples is now also AI enhanced.

Finally almost 2 years after everybody's being suuuper hyped about the AI possibilities Apple decided to join the club. Their WWDC (World Wide Developer Conference) in other words, hype machine to make investors happy. Was extremely successful. Just look at their stocks they went up like crazy!

You've probably already seen what Apple did announce:

• New iOS/iPadOS that copies Android
• Brand new Calculator that is now available on iPad and every other product. Key here is that a 2600$ device still (until Sep) comes without a built-in calculator app. You better have some pen and paper to do the math.
• AI Artifi... Apple Intelligence - the thing we care about.

​

Before we talk AI, let's talk about Apple first.

Why is Apple late?

Apple always joins the mainstream tech later than the masses. It is normal for this company to take it's time to think about the decisions the make. For example when the iPhone was introduced, there were already a plethora of touch-screen devices that were capable of so much more!

When the iPad was introduced - Tablet industry was extremely varied and there were tons of options to choose from.

Macbook Air - was not the first ultrabook on the market, but one of the most popular.

The most desirable products are built by Apple. Because every-time even if late, these products were simply the best (for the average user)

While the world ships MVPs and delivers unfinished 𝛼 or β products to the the customers, Apple has the budget and money to deliver their products once they are happy with the quality and user experience.

This is extremely expensive though and does not fit anybody. Only a handful of companies out there can afford to spend so much money on a product that a company doesn't even know if the users will use it, Apple Car, AirPower.

Why do the still do it like this?

Because this is the only way to deliver a high quality, well done, good product.

This approach is a legacy from Steve Jobs, and it has served them well. It is the only differentiator on the market.

Why we give our hard earned money to Apple?

1. Marketing and market fit. Apple always sells an experience, a status change and life style.
2. Product quality. If we talk Apple, the products are always high-end materials, specific finish, and a fantastic software experience.
3. Life Style / Social Inclusivity - If you own an Apple Product, you associate yourself with the people that own that product, and hence if you want to join a specific group, you can ease your access by purchasing associated things. Whether we need it or not is a different question.

What should we expect from Apple Intelligence?

Given the above mentioned information, we should expect a unique experience, we should see impeccable performance, we should see lightning fast AI.

Apple has previously patented an AI (Machine Learning Algorithm), that analyzes your screen and suggests and completes actions for you, and the best thing is it does this on device. Why is this good? well, it means there's no required internet connection; it will be done extremely fast; it will be very accurate; it can improve over time; it will adapt to your habits; It has lower security implications (since there's no network implied); less chances your data will be exposed.

Improved Notification grouping

What will be finally great will be notification prioritization. Due the fact that the AI will see which notifications you read first and pay most attention to, it will automatically select you things you care about, and silence shit you don't need.

Image Playground

Apple already announced that they have image generation on device, but frankly speaking it's kinda shit 💩, but hey, it's on device, and this means low latency, quick to deliver.

Text recognition and summarization

This will be available across all Apple products, and it'll be a killer feature. I would love this honestly, because now I'll be able to rephrase summarize and auto-complete content simply in my notes, without using a third party app like texti or chat - gpt.

It'll also be on device, and this means SPEED!

OpenAI + Apple = ❤️ ?

What is this all about? Well Apple doesn't have a trained model that can do so many things as OpenAI does, and on device models are not even close as performant and smart to accomplish the same tasks as a Large Language Models running on expensive hardware, that's why Apple can delegate hard tasks to OpenAI.

This means OpenAI is now a Business-to-Business SaaS. This also means that Apple Devices have now the best performing AI on the market, for free. This has security implications, yet Apple assures us that customers are going to be protected by Private Cloud Compute (their firewall) that will make sure to anonymize the data and always provide only the good things from OpenAI.

IMO this is simply a patch, for Apple until they develop and deliver a proper product that will be better than OpenAI. This Patch is there to last at least 1-2 years. I don't expect anything better earlier than 2025 or 2026.

Conclusion

As closing words for today I would like to say, that Apple did well, developers can have hands-on experience already. You can install latest beta OS and try it out yourself to see on any Apple device you have.

I honestly think this is going to bring so much AI into the mainstream! Ufff I love the future that is waiting for us!

That's it folks, see you next week ❤️️️️️️️

Happy Prompting!

​

Remember to invite your friends to subscribe at https://​newsletter.texti.app​

These weeks so many things happened it got me overwhelmed 😰. Felt like it's hard to stay on top of the AI trend. When things just get bad, breathe and worry not. Then you can see the clear picture.

Let's dive in.

The highlights:

1. OpenAI Chat GPT-4o​
   A new model capable of understanding voice, seeing things and flirting with you, with the voice of Scarlett Johanson. Note, the voice interaction is super fast, and it is literally a real time interaction. Best of all! It's FREE!
2. Google Gemini IO​
   Veo - Sora concurrent (that is still just a demo, hidden behind a waitlist)
   Imagen 3 - A very interesting image builder, that inspires me to change image generator for Texti app.
   2 Million Tokens Privately for Gemini
   Ask Photos - Ask AI to find you a story about your pictures stored in google cloud.
   ​Project Astra - the Star of the show! Multi-modal highly interactive demo, but this yet again just a demo 😅

​GPT-4o (omni)​

So OpenAI in their usual fashion have made everybody look like beginners. Google now looks like it's just playing around, and they really have no idea what they are doing.

One of the most fantastic things that OpenAI did, was they were just having fun with the AI model, and this means the product is enjoyable.

I truly believe that when a product is enjoyable, it's actually ready for people to use.

Live Translations - Spanish to English. The video does it all. It's fabulous and the speed of translation is acceptable. You could in theory rely on GPT-4o to get on a live chat with someone you don't speak the language.

OpenAI @OpenAI Realtime translation with GPT-4o OpenAI @OpenAI Two GPT-4os interacting and singing 9:39 PM • May 13, 2024 760 Retweets 4770 Likes Read 83 replies

​

Be my eyes - this is probably my favourite one. You can now rely on your phone as a guide for blind people. Happy to hear that tech is used for people with disabilities!

Blindness is undoubtedly an impediment in life, yet I've seen people excel in life regardless of it. When tech can help someone overcome a problem, it makes me think we live in a world where people aren't just selfish.

OpenAI @OpenAI @BeMyEyes with GPT-4o OpenAI @OpenAI Happy birthday with GPT-4o 9:39 PM • May 13, 2024 776 Retweets 4589 Likes Read 91 replies

​

Multi-modal example. Let's keep it straight, this is cheats level god. You are now not protected by anything, you can just show a sheet of paper to gpt-4o and it will know how to read it and how to solve it properly. Purely – fantastic 🪄

Also if you ever had issues to learn or understand something because there were no tutor around, then now you don't need to be afraid there's not going to be any possibilities. You can ask as many dumb questions as you want, no more shame in dumb questions. A robot does not need to care about feelings, it's all just pure information. The only necessary thing is a will to learn 👨‍🎓

OpenAI @OpenAI Math problems with GPT-4o and @khanacademy OpenAI @OpenAI Sarcasm with GPT-4o 9:39 PM • May 13, 2024 679 Retweets 4369 Likes Read 79 replies

​

There are many more examples of using GPT-4o in this thread on twitter, feel free to check them out!

An important thing to note, is that the voice of the assistant resembles heavily the voice of Scarlett Johanson, and she's suing them for that 😂. But that's a story for another day.

Anyhow, the speed of the interaction, the fact that you can now speak over to cancel and re-direct the assistant into the right place is freaking fantastic. It feels pretty close to a human that you can chat with.

​Google I/O "24​

To begin with, we need to make sure what google is doing. Google is literally throwing mud to a wall and sees what sticks. It launches more projects that it can think of, and see what project reaches a desired interesting number for them. If it doesn't go well, it's just going to be thrown in the bin. The google graveyard is yet to see a bunch of new projects.

Veo - this as presented by google a mashup of Video Poet, WALT, Lumiere, Phenaki. It is supposed to learn from all, convert it into super powers and bring into light a product you'd love. However, this is all same as before. Just a demo, with no proof. They say they asked creators to use it, but we don't know what really lies behind it. For now all we can do is just say we're waiting for access to check it out on ourselves.

​Imagen 3 - google's answer to midjourney, but I mean we all know it's not even close. The UI/UX is great though, we should definitely take notes from this and try running some improvements for ourselves. Below you can see a prompt that normally would run on a model and see the output. The UX of the tags at the bottom really makes a difference, it does inspire you to try out different styles and contexts of the image generation.

The results though are fine, not exceptional or extremely creative by any means, but the prompt is as simple as it gets as well. Also the a very important note is that fingers are drawn correctly so that's also a decent output.

Feel free to try it out by yourself here: https://aitestkitchen.withgoogle.com/tools/image-fx​

One of the highlights of the I/O was and still is Project Astra, this is as everybody says these days ground breaking, wow, new and impressive. A multi-modal capable interface-enabled AI that would be able to help you with answering questions you don't have an answer to.

The core principle being as simple as. You look at it, ask about it, and it'll give you the answer straight on point. Check their demo bellow.

It's is very important, this is simply a demo, it's not necessary a product that is ready to be launched. It could be all be just a recorded and edited video just for the sake of AI hype to keep the investors happy. Google is known to run such "experiments" 😅

Concluding today's newsletter, I just want to remind you. Our computers can now see, percept our voice, and respond in a fun and interactive way. If this is ain't the future I don't know what is. Enjoy the moment! ❤️

That's it folks, see you next week ❤️️️️️️️

Happy Prompting!

​

Remember to invite your friends to subscribe at https://​newsletter.texti.app​

This week Apple made official excuses for their ad. Apparently destroying a bunch of musical instruments under a press, that require hundreds of hours of craftsmanship to build and years to master, wasn't really a pleasant picture and didn't show their new lightest and thinnest device ever in the best light. At least they acknowledged their mistake. 👌

The highlights of the week:

1. ​8 Pixels to dominate them all - the new budget-friendly Google device.​
   AI enahnced, Good camera, good eco-system, and a fair price, what is there else to want? But is it tho? Google doesn't have flawless records. Their ads are pretty miss-leading (cough GEMINI cough 😷) so we need to get our hands on to see if they prove their promises.
2. Eleven Labs has broken another AI Music Barrier​
   They have tweeted about a new magical instrument (they didn't publish yet), that can generate 3 minutes long songs in different genres and can actually make it pretty decent.

​Google Pixel 8A​

Google has pushed the flagman features of last-year's phones into their little budget-friendly devices. Now screen calling, magic eraser and other things AI enabled features are on the lower end devices.

This year the new device has a couple new tricks up it's sleeve. New features like Best Take (see gif above), Audio Magic Eraser (see gif bellow), and Gemini are coming to the device.

As long as you can enable gemini extensions, you can ask your Pixel to, find flights and hotels for your next travel, or find youtube videos, or even ask it to compose a script for your next youtube short, almost like they integrated rabbit r1 in their phone 😅. Now you don't need to go to chat-gpt do that but rather ask your google assistant and it's going to do it for you.

This could be one of the best uses of AI-enhanced products yet. Daily driving a Pixel 6 Pro with the magic eraser, it was mostly a fascinating experience when removing my ex from all our travel pics. Note it was a bit of a hit or miss.

Given the Pixel's history, they got their trust with hard work, and the bar is pretty high, (even though they sell your personal data for money). Doing an oopsie is going to a be a very costly mistake. Let's hope it's not going to be.

Another feature I found pretty cool and honestly interesting is promoting consumerism at it's finest. It's called circle to search - it allows you to circle an item in any video and it will automatically search for it within online stores.

If you're here for the technical terms then:

• (8 square 😅) 64 MegaPixel main camera, and a 13 Ultra Wide Camera
• 13 MegaPixel front camera
• Tensor G3 Chip (what ever that means) but it is AI-enabled and that's only thing that matters I guess.
• OLED 120Hz display
• up to 1400 HDR
• Battery 4,492mAh (should be good for a day)
• 128/256GB of storage
• WiFi 6
• Price from $499 (of course taxes vary depending on your region)

Eleven Labs's AI Music Generator

Another day another another music generator. This time this comes from a company called Eleven Labs. This company dubs videos in any language using AI, simulating any voice (a lot of work done in the audio space).

On the 9th of May, @elevenlabs shared on twitter their new music generation results. It sounds very impressive really, just take a hearing bellow.

Personally I think the voice-over is really good, the words give it away and it kind of sucks, but other than that it's really well rounded.

ElevenLabs @elevenlabsio Here’s an early preview of ElevenLabs Music. All of the songs in this thread were generated from a single text prompt with no edits. Title: It Started to Sing Style: “Pop pop-rock, country, top charts song.” 7:52 PM • May 9, 2024 1027 Retweets 4424 Likes Read 307 replies

​

The progression in audio music generation, is advancing really fast. This is a crucial instrument that has to be considered when producing music in this day and age. It should be treated same way as synthesizers were, at first ignored and disliked by people, but later widely adopted.

Leaving that behind is just close-minded. You must also be aware that handcrafting is for a larger part a form of art, and mastering it takes a ton of perseverance and hard work - which automatically exclude a large pool of people that do not have the patience.

Just embrace the music generation now, adopt it, and expand your horizons, and iterate fast. Now you don't need to think and compose a melody, you can just generate it in a couple of minutes, and iterate with the styles.

Remember the future is bright, as long as you believe it, but also the future is now.

That's it folks, see you next week ❤️️️️️️️

Happy Prompting!

​

Remember to invite your friends to subscribe at https://​newsletter.texti.app​

25 is a very lovable number, I don't know why but I meet this number a lot in my life. It's a quarter, it's a golden age, it's a threshold for DiCaprio's girlfriends, it's an odd yet incredible number. Kind of love it. How do you feel about it?

The highlights of the week:

1. Rabbit R1 has shipped and it's looking grim 💀​
   People got their devices and it seems it's kind of shit to use. It is an android app on an android device who's battery dies in 2 hours, and somebody already unpacked it and installed it on an iPhone, even modified it. 😂
2. ​Nothing Buds with integrated ChatGPT​
   The title says it all. You can ask questions and interact with ChatGPT with your earbuds. Neat! But is this innovation or just another gimmick?

Rabbit R1 has shipped and it's looking grim 💀

Last week the Rabbit R1, started shipping to people, and everybody started unpacking what's locked in that package. It is clear that this had an ambition, but from day 1 everybody understood, it has to be an app. Yet nobody's gonna buy an app for $200. So now that people got hands-on it, what does it feel to own one?

1. The battery dies in 2 hours. Similar to the Humane AI Pin the device dies in basically no time. You have to charge it multiple times a day 💩
2. The latency is tremendous. I remember in the demo last year, their demo was really cool, it was snappy and responded almost instantly. From the reviews it seems that the device takes it's time to reply ⏳. That was a killer feature, and I thought they're going to put a model on device, and use that instead of on cloud query evaluation. That'd justify the device.
3. The control implementation is really bad as well. I mean the idea is suuuper cool, I like an analog control wheel that lets you access quickly your desired place instead of wiping your screen. I find this brilliant!
   R1 did not deliver 😓. Their implementation is simply incomplete. It is slow, inaccurate, sluggish and simply painful to use. If you've ever used an iPod Classic, you know how good that wheel was. That's what this device should've had.
4. Their LAM (Large Action Model) they were bragging about, is as raw as the fish swimming in the ocean 🍣. It hardly does the activities it was promised to do. All it can do is navigate a couple of apps, if it doesn't hallucinate.
   Personally I think, this is barely a beta product, more like an alpha, yet to be given to the testers.
5. Lastly a developer from NYC, @WillHobick, has unpacked the android app that is running on the Rabbit R1, and installed it on an iPhone, and on the web, he then even customized the colors and added more functionality to it, to show how it's done. That's funny because it took em 6 months to deliver a product, and it took 1 week to convert it to a fully functional cross platform app. 😅Investors money spent poorly, running on top of the AI Hype. 💸

Will Hobick @WillHobick Update on r1 rebuilt on iOS ✨ Added haptic feedback, new buttons, updating image functionality to better match the device, image history and more! Also, I've changed the color to blue 💎 2:32 AM • May 4, 2024 39 Retweets 844 Likes Read 29 replies

​

Nothing Buds with ChatGPT

If you didn't know, Nothing is a company that builds consumer tech, they build 2 phones so far, and this is their second generation of ear buds. The market for consumer tech is pretty tough these days, and it's really hard to stand-out in the crowd. So they're always hunting for something.

Now look, I understand I just spilled shit over an AI product, HOWEVER, I think this is the right way to build an AI product. First of all they don't charge 200$ for an app, but a fair price of ~$100 for their bluetooth buds with noise canceling and a cool looking design, with an extra feature of Voice activated ChatGPT.

In my opinion, AI products should really augment and improve already existing tools but not try and replace them completely. It does create niche products but hey, that's

I know this is not that futuristic, and is merely different to a voice assistant, but hey competition makes humans progress.

That's it folks, see you next week ❤️️️️️️️

Happy Prompting!

​

Remember to invite your friends to subscribe at https://​newsletter.texti.app​

Fun fact: 24 is what is used to measure 100% gold - like our today's newsletter.

The highlights of the week:

1. ​Amazon's Maestro​
   Clearly a RIP-off of the Apple's Genius playlist creator, that targets Gen-Z and later because it's fully emoji friendly 🤪.
2. ​Meta's New AI Glasses​
   Glasses that augment reality, they record videos, stream video calls, answer your questions and are AI multi-modal enabled. Also I really want them 😅

​Amazon Maestro​

It's a Genius move because it's a RIP-off Apple's magic playlist creator and it's called Maestro. It's powered by "AI" - not sure we can trust Amazon on this one 😅, yet we kind of have to.

Anyways, it's magic because it can create a playlist based on this query "😭 and eating 🍝" or better "🍆 + 🍑".

I mean just by the prompt, they're showing that they are targeting this music playlist generator to people under 20 years old, because it heavily relies on emoji and the keyword boyfriend is persistent.

I understand stealing kids from spotify and moving them to Amazon Music, could be another Genius move, as it could yield long term customers, which yields a very valuable customer acquisition. They want this piece of cake the social media has. young blood that has has access to tons of parent's money. So they're definitely pushing through.

The feature is already live for handful of customers based in the US, and it is yet to come someday to the rest of the world. Amazon music has 13% of the global market, but they want waaay more. That's a battle uphill that Amazon is yet to fight.

30 seconds break

Here's a video of showing how to hack a visual based interface. Literally a person with a T-shirt with a stop sign, would cause AI based autopilots to stop the cars in the middle of the road. I find this super funny and frightening at the same time 😂.

Itamar Golan 🤓 @ItakGol Visual Prompt Injection 💉🛑 IRL 2:34 PM • Apr 26, 2024 387 Retweets 2078 Likes Read 31 replies

​

​Meta's New AI Glasses​

This week Meta has launched a new ad campaign for their glasses, because they've got an update.

They now support multi-modal capabilities. (Multi-modal means they can process visual and audio information), but the ad is suuuper cool! Now I really wanna buy em. They also promote a retro like style vibe and somewhat very stylish.

Enough back to tech. They can record videos, stream video calls, be interactive with you while you do something, take pictures, explain the world around you.

I feel like this is the best most comfortable, and accessible continuous augmented reality we currently have. Given you will have to have constant internet access to use them, yet still this is a very neat package to be able to consume.

Dang this feels like an ad 😅, but I mean I really like the marketing and package presentation they have. I should definitely buy them and get disappointed so that I could be more critical on the feedback.

Anyways what I wanted to say, well if this isn't the future. You can talk to your glasses like it's an extension of your eye. It can see, react and act, as if this is nothing but normal. We live in the freaking future enjoy it.

That's it folks, see you next week ❤️️️️️️️

Happy Prompting!

​

Remember to invite your friends to subscribe at https://​newsletter.texti.app​

Today we start-out with a random cool fact, did you know that 23 in Japanese is pronounced like ni-jū san, which is literally the name of the car company Nissan 😅

The highlights of the week:

1. Meta released their Llama 3 🦙🦙🦙 Models ​
   Trained on 700% more data than Llama 2 🦙🦙 and 400% more code! It has also now has a prompt input of up-to 8k tokens, which gives them the unique advantage to brag about how cool it is, and that it's better in synthetic test than the competition 💪
2. Amazon's Just Walk Out isn't really AI-Tech - a rather HI (Human Intelligence)
   Amazon Just Walk Out stores where you could grab food and go through without a checkout, which they told everybody is AI - powered, wasn't AI all along, but a rather 1000 Indian workers that were doing the checkout for the customers. 🤡

​Llama 3 🦙🦙🦙

What is this lama all about? Well Facebook, has been launching open-source models for a couple of years now. This week they announced a new gen-3 AI model called Llama 3.

- Why should I bother?

You shouldn't bother 😅, you just should know that you can use it at meta.ai, that is if you are in the US, otherwise just plug-in your favourite VPN, and the floor is yours.

- What can it do?

It can do the same thing as others do, but just better, or at least that's what's Facebook sayin.

- I like when my ice-cream 🍦 has different flavours!

Worry not! Llama 3 has 2 flavours too! It comes in Llama 3 70B and 8B, which literally counts their parameters, so 70 Billion parameters and 8 billion parameters. These parameters shows how much pre-trained data they are using, but this is a little less transparent, so think of it as magic 🪄.

- Enough! I want to download and test them myself!

Of course, just share some of your personal details or (fake ones + 10MinuteMail) with facebook and download them all right here.

​

All in all, I think this is progress, and as a developer I'm super happy this model is open source, because it allows people without the expensive gear experiment and play around with the AI capabilities!

Amazon's Just Walk Out - is a fluke!

If you are in a few countries in the world where Amazon has it's own grocery stores powered by Just Walk Out Services, then you might've ever wondered, how does that work? It's magic!

Amazon says it's powered by AI, cameras, sensors, bluetooth and other wizardry. However, it ain't that magic at all, it seems that it's powered by 1000 Indian Workers in offices that check-out for you on your web portal 💩.

I don't know if this needs extra commentary honestly, it's a fun fact how companies are faking out technology, and how many times you either get caught, either get away with it.

Isn't this an example of fake until you make it? An example of never stop hustling even if you're a multi billion dollar company! Just make that shit work, make it a good product, and fix the details later, if you can afford it. If it didn't work out - it'll be easy to get over it and move on.

That's it folks, see you next week ❤️️️️️️️

Happy Prompting!

Remember to invite your friends to subscribe at https://​newsletter.texti.app​

Today we get straight to the point.

The highlights of the week:

1. The Humane AI Pin is here.​
   In a couple of words, this device wasn't beloved by the consumers. The battery dies within 2-4 hours, it's always hot, the interface is pretty hard to use, the lag is unreal. However the build quality appears to be fantastic. The price is really high.
   ​
2. ​Udio the new music AI generator.​
   You thought last week's Stable Audio was cool? Duh. New Udio is even cooler! The results you get from it, are close to unbelievable. It has full text verses, and the music pretty genuine. I still managed to make it hallucinate 😅

The Humane AI Pin is here.

5 Months ago in November 2023, the world has seen a major announcements. An ex-Apple team said they will produce a product this world, have never seen before.

They announced a Pin-like form-factor, which clips to your clothes with a magnet, and is a fully autonomous AI-powered device. It has it's own internet connection, it's own camera, it's own audio controls and it's own battery, and most futuristic a palm projector to interact with it like a screen. Promising! 🥳
All of this at just 700$ + 20$ monthly subscription (not so promising 😢)

Well they promised - they delivered 🚀

This week first deliveries started handing in. People started receiving their much awaited devices. Youtubers are obviously the first ones. In my recommendations MKBHD was the first one. From what I've seen here's my opinion based on the influencer's opinion.

To put it lightly this is just an unusable device in it's current form, and mainly because of the following reasons:

1. The time to response is slow 🐢 - Every question takes about 5-10 seconds to answer, given that you have a stable internet connection. That's ages to wait
2. The battery dies super fast 🪫 - anything that stays under 6 hours for the modern day is literally unreliable. As reviewed by MKBHD this dies in 2-4 hours.
3. Lack of integration with consumer's information ℹ️ - in other words the device is truly stand-alone and doesn't want to integrate with any apps and doesn't know anything about you. It lacks the purpose of an Assistant - it does not know anything about you.
4. Less than ideal interface 📺 - the way to communicate with other people through the AI Pin is just ungodly unusable.
   Adjusting details for your requests is literally impossible. How many times a day do you correct a phrase you write as a message? With an AI interfaces that's hard to address, because instead of clicking a button, you need to explain which part of the message needs to be corrected.
5. Projector 📽️ - is good only in the shadows
6. Lack of Privacy 🔐 - the feedback of the pin is done only via audio or Projector, neither of which are exactly private. Also projection is literally in one place, so you always need to keep your hand in a very specific position.
7. Always on fire 🔥 - literally warm to touch all the time due to the wireless charging.
8. Potentially a perfect product to be stolen by thieves 🤡 - simply take it and run.

Apparently the only good engineers are hardware engineers. Because the hardware feels great in hand and well made.

The AI Pin will only be usable when it will manage to have the AI on device, and process your requests within milliseconds, everything else, is just a fluke and a beta test, to show-off.

Well yet another product living wasting investor's money. 💰

​Udio​ the new music AI generator

Another week another music generator.

I was expecting a lot of Videos for this year, but look who's first, the Audio scene is getting better by day.

This week's highlight it Udio. All I can say it's freaking incredible. It has bugs of course, but it's in beta and it is sooooooo goooood! I believe this is by far the best Audio Generation tool we've seen so far.

Of course I couldn't stand aside and generate myself an audio track. This newsletter premieres Dreams of Texti-App. Listen now on twitter or download the track right from this newsletter, add it to your Spotify playlist 🎧

I'll be honest, last week's Stable Audio was pretty cool, but it was lacking vocals. This week things look completely different. Like the quality of audio is nuts - and it even rhymes very neatly. I like it a lot.

Fun fact about this startup that's supported by will.i.am, yes – that one will.i.am the artist who apparently has embraced the power of AI and instead of fighting the revolution – joined it.

That's it folks, see you next week ❤️️️️️️️

If you liked this edition, please remember to invite your friends to join our family.
I really appreciate your subscription and keeps me motivated 🔥

​newsletter.texti.app​

Black Jack - that the mood. Apple is gambling with people's desires. And I found a really cool audio AI, which you should and have to try! Let's get into it.

The highlights of the week:

1. Apple rumors
2. Stability AI Music

Apple is rumored to explore home robots

This year Apple officially dropped their project of an Apple Car 🚗, that was meant to be something out of this world. An electric car that would've costed like a house in Maldives.

My uneducated guess would be, they took too long to make the car, and current electric cars market is a little down, so continuing to pursue the project is a little meaningless.

Now since they already have a massive team that used to work on the electric car - which implies hardware engineers, the only logical follow-up is to pour money in Robots 🤖.

With the AI hype pumping all day every day 💪, which enables robots to reach quite a new level of highs, and frankly speaking Apple being on top of the hardware industry, a robot is nothing more but a waiting game.

I personally can't wait until Will Smith is gonna save me from the robot world domination... oh wait iRobot is a trademark and they cannot claim it back! So Will will not save us from the iFuture.

That said, Apple is a trillion dollar company so they can just buy out the iRobot corporation registered in the US and everywhere else in the world.

Attention! To make sure this becomes a reality please go and buy yourself, and every member of your family an iPhone, and subscribe to texti newsletter! ❤️

​Stable Audio by Stability.ai

TLDR:

One more Music Generator 🎼, but it can generate up to 3 minutes long, and you can get a commercial license. Also it can generate music based on your lame beat box.

Long Story:

This is nothing new, we've talked about music generators already multiple times, but hey, when there's a new kid in town, we always have to try it out see what it can and what it cannot. 🧐

This time this a brand new music generator from Stability.ai, these guys are pretty cool and they are on pace and keep up with the big players.

Their Stable Audio generate pretty good results, most of the time 😅, but some are pretty shit 💩. Hey! give it some time, it'll get there. Bellow you can see an example of the a song I've generated, I'll be honest I just used some of their templates, without being to bothered of writing a prompt myself.

Texti App @TextiApp Wow @StabilityAI just launched their new toy: Stable Audio. You can generate lengthy music, compared to other tools like @suno_ai_. Best of all it was trained on music from @AudioSparx so nobody should be crying about it. Feel free to try it out on your own… https://twitter.com/i/web/status/1776145219248914766 10:9 AM • Apr 5, 2024 0 Retweets 0 Likes Read 1 replies

​

Pretty cool huh? But perhaps the most unique feature I haven't seen so far is the following. They can generate music based on what you sing.

Now you don't need to be afraid that you cannot sing, because all you need to do is just pass it to an AI, make a coffee, and listen to a brand new song to hit the charts in tiktok. I mean isn't this the dream? Popularity made with AI. Fucking awesome!

Tip of the day

Sometimes I want to explore things I don't understand much about. Like music. for me this a pretty unknown and unexplored space, I mean love listening to music, but their jargon is what I'm unfamiliar with. To make sure I know how to hit the right spots I use AI.

Here's an example:

I'm certain that texti doesn't know anything about Stable Audio (it was just released) so it doesn't know how to write the prompts out of the box. However I have an example prompt from Stable Audio, I just explain it to texti, so it understands it's task better.

And here's the result:

​https://stableaudio.com/1/share/de72a13e-32a3-4a5c-9327-77f7295024f3​

​

I mean I'm using AI to generate things for AI that generate AI. 😅

If you're not yet convinced that we live in the future, I don't know what will convince you otherwise.

That's it folks, see you next week ❤️️️️️️️

Happy Prompting!

Remember to invite your friends to subscribe at https://​newsletter.texti.app ❤️

This week's a jubilee! We're already on edition 20! Isn't this wonderful? To Celebrate this I would like to give you all a free month of Texti.app, text and image generation. Just use the code: TEXTINEWS20 . Only the first 100 people will be able to use it, so don't wait, got get it now!

The highlights of the week:

1. ​Infinity AI - deep fakes accessible to everyone.
2. ​Tavus.io​ - video templating of the future

​InfinityAI will make you deep fake!

TLDR:

​Infinity AI a company that believes that a couple of writers can win an Academy Award in the next 10 years. They generate videos based on your own script. And you can try it out yourself in their discord. Super cool stuff.

Long Story:

I just bumped on twitter on this cool company, these chaps are actually trying to make a reality the thing I said a couple of newsletters ago.

They are aiming to train an AI that will convert any script into a fully featured movie. With a skilled editor – this is should be a reality. We've already seen Computer Generated Paul Walker, and other CGI personas, I don't see why we should stop at CGI if we can enable AI to do it. AI is the ultimate human like parrot 🦜.

The above videos are built with Infinity AI in their discord channel. I would encourage you to try it out, as well. it's really fun and simple, and for now this is free. So grab it while it's hot.

I believe their mission is going to be realized at some point, and I think they deserve success.

I mean just with this tool you could generate a full interview between any of the following people. Have fun 😁

​Tavus.io - video templating of the future

TLDR:

A personalized AI video generator, which allows you to address to a specific person by name, however it can automatically take a list of say 10 000 people and generate a video specific to each of those people.

Long Story:

Today I found this Y-Combinator startup that is building and training their own AI. What it does is pretty simple, let's say you have a newsletter like this one, but it's in a video format, now you want to address to each of your subscribers by name in particular. How do you make that happen? Normally you'd make the video, and then edit out where you call someone by name, and edit video by video to make it personalized. Obviously this is actually trouble some, nobody does this, for 10 000 people that'd take an enormous amount of time. Well this company aims to automate this with AI. They train their models on your personalized face and then all you do is just give it names, and it'll change a specific part of the content. And voila it's done - automagically.

Now this cool but the moment I saw this, the first thing that came to my mind was the Sci.Fi movie Minority Report, starring Tom Cruise, from 2002. I remember there was a scene in the movie where Tom Cruise was on the run, and there were massive screens all over the place, that ran extremely personalized ads, as in "Hey John Anderton, it looks like you could use a Guiness right now." and it freaked me out so much back then. I was like wtf this is sooooo cool, but soooo frightening.

And it seems that we are finally there 🥲

PS. If you don't know what I'm talking about take a look at the following clip from the movie:

​

​

In conclusion I would like to say as usual, we see how the world is changing and evolving just in front of our eyes, it really means nothing more than:
​The future is now, and we're living it!

P.S if you like this newsletter and are you enjoying it please recommend it and share it with your friends. Your support and encouragement helps me out tremendously! ❤️

That's it folks, see you next week ❤️️️️️️️

Happy Prompting!

​

Remember to invite your friends to subscribe at https://​newsletter.texti.app​

Another week another and the hype is still on. AI is everywhere! Wherever you look you will find AI. I remember a time when Bluetooth became suuuper popular, and people were adding it everywhere. I feel this is exactly the same. I mean any hardware has some sort of AI integrated. So let's start with something not really AI for a change.

The highlights of the week:

1. Texti now has image generation!
2. Neuralink has made a human telepathic!
3. Sam Altman goes on a ride with Lex Friedman

Texti can now generate images

It was long awaited, and I wanted to make this happen for a really long time. Now however you can finally generate images directly in texti, for now this is going to be available only in the web platform. To try it out feel free to just go to your account on the web (https://texti.app/user/profile/images) and start generating images.

​Neuralink demo-ed their pacient 0 abilities

TLDR:

Noland Arbaugh, Patient 0, has been shown on a live stream on twitter. He showcased how he was playing chess, and how we was beaten by the computer (obviously). I mean he is indeed the first telekinetic human.

You can see the entire stream with the patient here: https://twitter.com/i/broadcasts/1ypJdkXjaLNGW​

Long Story:

Noland Arbaugh, Patient 0 of the company Neuralink, has live streamed the his usual life of a human with telekinetic powers. Not all of it but his usual session of losing a game of chess to a computer.

He described the learning process, where he learned how to move the cursor from point A to point B, and how marvelous this is. He funnily makes jokes about himself being the actual Professor X as he's able to control things with his mind.

Noland, seems like he loves playing Age of Empires, and he said the first time he was left with it to himself, he played for as long as the battery in the telekinesis device lasted, 8 hours 😅. Man is given the power to control computers at a distance, and first thing he does, is play computer games until 6 am. Men be men, we never grow up 😂. Sorry ladies.

Jokes aside I think this is marvelous 🪄, we live in a time, where we see characters from comic books become a reality, where people are enabled to control devices just with their mind, enabling the person to be able to work and be an integral part of the society, removing the disability almost to 0.

With a high enough demand, eventually people will be able to drive cars with their mind only, unblocking another major impediment for people with disabilities. But for those without any impairments – imagine how would formula one look in the future. In the 1990s we played mario kart on our computers, in 2060s we'll play mario kart in real life, just using our mind! 🤯

Sam Altman & Lex Friedman

TLDR:

2 hour long chat between Sam and Lex. Where they talk about what will GPT-5 be, the grandiosity of AI, how google isn't enough of a big fish for OpenAI, the even when he was fired from OpenAI, and then reinstantiated, how Elon wanted to buy them, an keep OpenAI as another child company. However make sure to play it at least at 2x. They talk as if they meditate really.

Long Story:

Oh boy, a really long interview. I'll try and cover the most important parts. But before even go into the specifics. I must say Altman thinks extremely global, the scale he thinks it is vast, and enormous. Just give you an example, he said, "replacing Google with AI, is not large enough for me." The scale he thinks at is just grandiose, but also take this with a grain a salt 🧂, as this means extremely dreamy and low fidelity – basically no specifics.

GPT-5

Sam says that GPT-5, must be a marvel, and that it has to surpass GPT-4 same way as GPT-4 surpassed GPT-3. He also neither confirms nor denies that GPT-5 will be out this year, but likely not. Money and focus is currently thrown at other things like Sora for example.

Elon wanted to buy OpenAI

As you may know from our previous post, when I was sharing my thoughts about Elon suing OpenAI for their name, we learned that Elon was among the founders of OpenAI, but what we did not know is that at some point in time, Elon wanted to take over the control over OpenAI and make it another sub-company like Tesla or SpaceX. Don't get me wrong this are companies with an extremely bright future and they already have a massive impact on the world, but OpenAI on it's own has already impacted the world way more than all above mentioned combined.

Expect great product launches soon

When asked about incoming launches, Sam obviously hasn't said given any specifics however, he said that there're big and exciting things coming, and that in-house they have already massive things done. It's just a matter of polishing until they arrive to the public.

Full Interview:

Conclusion

There were a tons of things discussed, but just to re-cap, this week we saw the first human control a computer with his mind. we saw Sam talking about the futuristic possibilities of AI and how much more we have ahead of us. I mean isn't this a marvelous time to be alive! We live in the future! ❤️

That's it folks, see you next week ❤️️️️️️️

Happy Prompting!

​

Remember to invite your friends to subscribe at https://​newsletter.texti.app​

This was a beautiful spring week, it blossomed beautifully and it was definitely the AI sprint week. Mid journey improved their face generation algorithm, and now it doesn't just generate same facial expression all the time. But the most important parts are the following:

The highlights of the week:

1. Devin takes the software engineer job
2. Figure 01 is now able to talk and interact with humans

Devin from Cognition labs, promises to take you job

TLDR:

Sorry this time you'll have to go through the entire story. I feel this is a pivoting point in our history.

Long Story:

Everybody's been talking about AI taking our jobs. First writers, then artists, now software engineers... 😭

A company called Cognition Labs, just announced their first AI Software Engineer. According to them it's the most capable of all of them so far, beating existing models considerably by about 100% (more on this later). It has it's own terminal, browser, code editor, learning abilities, and a time-line to follow through it's progress. Veeery impressive!

In the following demo we can see how Devin, went through a blog post, and learned how to generate images hidden in plain sight, and generated some images there.

I mean this looks extremely promising. The media went:
​Omg omg omg, software engineers now you gonna suck dicks as well, death from your own creation! Hahahahahahaahhahahah *cough, cough* hahahahahhaha!!!111

Now, let's be real, I'm a software engineer, and I would like to go one by one with this one.

1. Am I afraid I'm going to lose my job? ​
   Not really, what Devin does is extremely technical and an uneducated person wouldn't really understand how to fix it.
2. Will AI replace my job?
   ​Eventually, of course it will replace my current job. However, that said, I think the future software engineer is a micro-manager of multiple Devins.
   To make the best analogy I can to the real world, is System Administrators manage multiple Servers and environments at the same time. Servers work on their own, but from time to time they glitch out, and you need to come by and fix em. This is what a software engineer of the future will look like.
3. But it beat existing models considerably by about 100%
   ​Which means that it can currently solve about 13% of the existing problems in the given bench mark, so in the other 87%, a human will still do better.
4. But it can pass job interviews! and I cannot! 😭
   ​Job interviews often rely on general knowledge and algorithm solving. And well machines are the best at solving problems based on a step by step instruction. It's also cheating, because AI has access to internet at all times, while you as a human must remember things by heart. It's not really an apples to apples 🍎 != 🍏 comparison.
5. But it can already complete jobs on upwork!
   ​There are very trivial jobs in upwork that can be completed literally by anybody, that can dedicate 2 hours of their life in understanding how to accomplish the job. Look at it from the bright side – devin will create your upwork profile and then you'll get jobs because it's going to be easier to apply.

Cognition Demos:

​

Cognition @cognition_labs Today we're excited to introduce Devin, the first AI software engineer. Devin is the new state-of-the-art on the SWE-Bench coding benchmark, has successfully passed practical engineering interviews from leading AI companies, and has even completed real jobs on Upwork. Devin is… https://twitter.com/i/web/status/1767548763134964000 3:50 PM • Mar 12, 2024 10650 Retweets 44735 Likes Read 4158 replies

​

Simply by looking at the promoted demos from congition labs, you can draw some conclusions.

1. Software engineer jobs are going to change drastically, and the industry will adapt to it fast. Devin is just the beginning and to me it's more of a copilot on steroids to which you can give stupid tasks, and focus more on productive work rather than monkey job.
2. They've shown only what it can, purposely hiding what it cannot, just to jump onto the hype train, misdirecting people's attention from the what it can't do and bringing more attention from investors to pour money in.
3. Probably one of the most important thing is: You have to give pretty technical and precise instructions. This one is a pretty big deal. Because creative people, often lack technical and precise knowledge, they know what they want, but they don't know how to explain it in details, and that's the Achilles' heel, of general purpose large language models.

While we're closing the gap there's still a long way ahead.

iRobot movie scenario is closer than you think

TLDR:

A company called Figure.ai, just demo-ed their own robot, that can talk and interact with humans, using the power of OpenAI, using multi-modal capabilities, talk, see, interact and remember! Suuuper Impresive.

Long Story:

This week a company called Figure.ai, showed what they have achieved with that $675 Million raised from Nvidia, Jeff Bezos, OpenAI and Microsoft. Their robot was able to:

1. Distinguish objects on the table, by using a 10Hz (10 images per second), camera for eyes.
2. Percept and follow audio instructions given via microphones in the body
3. Interact with the objects on the table, in a very accurate manner, using 100Hz (100times per second) signals given to the robot arms, and body.
4. Mimic human body behavior when moving, to keep it human friendly and approachable.
5. Remember the context of the whole conversation through-out.

This is very impressive and probably one of the most advanced Robots we've seen so far! Movement accuracy and control is just soooo good, i'm very impressed! But it's also suuuper slow to answer, feels like a sloth from Zootopia.

Technically it has a battery that can last about 5 hours, so must be always close to the socket, and it's still a prototype, still a long way to go until mass-production, but really promising at the moment.

I cannot stand the similarities it has visually to the robots from iRobot, Sci-Fi movie.​

We stand at a pivoting point in our life once again, on one side, Software Engineers are about to go extinct, on the other side, Robots are to take over another set of jobs:

1. Cleaning services
2. Construction workers
3. Mining
4. Agriculture
5. Anything that can be done physically following a simple procedure.

The future might look scary, but don't worry it isn't. You as a human will still be there to control the robot most of the time, and as in the case of software engineer jobs, instead of managing a bunch of software bots, you as a human will have the power to control a bunch of actual robots 🤖! You will have your own empire of robots that will execute your orders! And as such will grant you an edge over everybody else.

Just imagine if you could work in the harshest conditions on earth and all it's gonna take, isn't to loose a human, but rather replace a motherboard?

You could make oases 🏝️ from places people used to think were uninhabitable before, because you could just throw in a bunch of robots to handle the heavy lifting, and you'd come over and collect the harvest!

I truly believe the future is bright because technology is here to serve humans, and not to overtake them, it might not look like this at first glance, but take one step back and you will see a greater picture!

What a time to be alive! The future is now! ❤️️️️️️️

Happy Prompting!

​

Remember to invite your friends to subscribe at https://​newsletter.texti.app​

Women's day has passed, have you congratulated your beloved ones? If not? go ahead get some flowers, or at least generate some and let's get into this week's highlights.

The highlights of the week:

1. The social media is about to get blown up!
2. Google trains their AI using radio

​InVideo.io

TLDR:

This AI tool can generate you a fully fledged AI video, using stock footage and generate videos by any topic. Check the preview! To see a result. It's also free for now

Long Story:

We're almost there! A startup called InVideo, is now able to assemble and write a script about a topic of your choice. It guides you nicely throughout the process to understand what platform you are targeting.

Based on your input and instructions it will assemble a script, consisting of several chapters. Each chapter is built with some stock footage and hoping it's based on your script. Granted it does not really generate video so it's not Sora or even close to it, but it uses stock footage so the vids are not really shitty but rather good.

In the case the AI went nuts, and generated a bunch of gibberish, you can also manually edit your scripts, delete or add entire chapters. You can even replace or remove the stock footage that your video looks much nicer in the end.

Generally I think this is already high up there, with and moves the needle further. You can really make things happen with this.

Note: You can also add your own videos instead of stock footage, but that's not fun anymore. 😅

​Google is training it's AI with Radio​

TLDR:

Google just filed for a patent to make it's AI train on radio streams, this will make the AI learn natural voice, and has likely unlimited voice audio to learn from. I find it ingenious!

Long Story:

Google found it's unlimited resources of audio data, it's free, it's streamed live non-stop, you can learn natural voice, you can learn rare languages, you can learn all the languages, all you need to do is just listen to radio.

Google just filed for a patent where they describe entirely the procedure of training their own AI models on radio data. Reminder if you want to know why exactly google needs audio sources in the first place you can read a super simple explanation here.

I honestly find this pretty ingenious, you have all this raw data of voices, which are streamed over internet a lot of the time, and all you need to do, is just feed them to an algorithm that's gonna break it into computer understandable and distinguishable state. What computers understand are numbers, so that's exactly what they are going to do.

Audio is a audio wave, which means it has a unique fingerprint. All of these fingerprints can then later be grouped and aligned into a database.

Now with tons of trial and error, until this is polished up properly, Google will eventually have a vast source of audio profiles in most languages.

AI will be able to synthesize any text into audio, literally speak as a radio broad caster 📻 🎙️

The future is now ❤️️️️️️️

Happy Prompting!

​

Remember to invite your friends to subscribe at https://​newsletter.texti.app​

Today into the highlights of the week, something I was not expecting to happen that soon! This thing is really incredible and I was really hyped when google announced it, but they lack behind their product managers.

The highlights of the week:

1. Next Generation Text2Image AI is finally here!
2. Elon Musk Sues OpenAI 🤯🤦
3. Make it sing!

​Fast SDXL AI​

TLDR:

This is gonna blow your mind 🤯, it generates images as you type!!!

Long Story:

I don't think I can type a better story that the TLDR, I mean it freaking generates the image on the fly! Yes it isn't perfect by any means, but holy it's done on the FLY!!!
​Just 3 weeks ago I mentioned and showed google's video demo about this, but now this is a hands-on experience, where you go to: https://fastsdxl.ai/ and start typing.
Feel free to check-out the following video. It is not sped-up it, it's real time.

If you think why does it impress me so much? Well there's a simple reason. And that this eases the learning curve dramatically. Instead of waiting for image to show up for a minute or so, you just start typing and voila the image is here!
​
Damn this is impressive, I just can't get over it! Please try it out! https://fastsdxl.ai/​

Elon Musk is suing OpenAI

TLDR:

Elon is not happy OpenAI is not so open as they seem be so "open". And I think Elon is correct here.

Long Story:

I didn't know this, but apparently Sam and Elon are good friends. You see back in 2015 the company was founded by Sam Altman, Greg Brockman, Reid Hoffman, Jessica Livingston, Peter Thiel, Elon Musk. These key people agreed that AI technology must be open source, and free to access by everybody. It also must have enabled the entire tech community to contribute and improve it.
However everything changed in 2019/2020 when they realized they sit on a massive well of gold 🏆.

Elon Musk has had his own activities since so he wasn't really a part of the gang actively developing it, but he was donating money to OpenAI to maintain it's activity and development. A grand total of $44 million, was donated by Elon.

Since Microsoft has taken over OpenAI, it stopped being open, and more of a closed source company, that just shares an interface to use. This obviously makes Elon and the founding fathers a little unhappy. Because a mission once set, has dramatically drifted in a completely opposite direction.

I don't always say this, but I think Elon is correct here, because once an honorable mission is now a fluke of the past. Open source will eventually produce way more advancement and faster advancements than any closed source company, it is granted a little more dangerous, but everything has 2 sides, and a large gray area in the middle. Open Source community means no barriers to enter and try tinkering with it to deliver an "accidentally" fantastic output. I truly believe a company like this must be open source.

No matter how many employees you have, a community of 8 Billion people is still larger and doesn't need as many managers to build the future.

​AI Sync Lip​

TLDR:

A new research was just posted out, where an impressive conglomerate of Image2Image algorithms put together a model that can transform an image, to a video, where the face in the picture, could do sync lip the music in the background.

Long Story:

This week on twitter/X I stumbled across this incredible LLM that's capable of doing Image 2 Video and then Video 2 Sync Lip, based on an audio. Holy next level of deep fakes. I mean this is sooo cool!

If you access the research page, you can see an in depth explanation on how this is done. But in a nutshell what they do, is:

1. Take an input image.
2. Generate a bunch of variations of that image, with possible face gestures.
3. Convert audio to a time line, of words and music, so that you can match the face gestures.
4. align face gestures with the audio timeline
5. Mash them together
6. Output a video that leaves you out of the water.

I honestly believe there's a ton of potential in this. It is very similar to last's week Gen-Z math tutor, which made it incredibly cool.
Some ideas that pop-out immediately:

1. Missed Connections: Imagine you want to talk to someone from the past, who has sadly passed away. You can just add a voice over, and the AI will take care of the rest, it'll make your relative almost alive.
2. AI is sound: You've heard of fictional characters before, now they will actually be able to communicate. Imagine a twitch stream with a completely fictional character, where an AI plays a game and humans just watch it.
3. Fictional Impressario: Absolutely fictional Holywood actors could now become a reality, including stunt doubles that were used before, now could be a thing of the past, and nobody gets hurt during a movie scene.
4. Security Risks ⚠️: this is kind of the other side of this power, you can impersonate sooo many people. I think we'll eventually have to have keywords instead of "hello" to make sure this person is actually a person, and not an AI, or somebody impersonating you.

Oh the world of the future is so different! It's so exciting!

That's it folks, see you next week ❤️️️️️️️

Happy Prompting!

​

Remember to invite your friends to subscribe at https://​newsletter.texti.app​

This week is the last week of the second month of the year! 🤯 16% of the year is now gone, how does it feel? Google fucked up already thrice in this time, so it can't be that bad.

Before we proceed to News, if you were ever interested what is AI and how does it work, in a suuuuper high level approach. Please feel free to take a look at our FAQ page, where you can have a grasp of it.

If you have more questions you're interested about please feel free to send them at inquiry@texti.app, I will gladly take my time to answer them. Now back to the news.

The highlights of the week:

1. Google is Racist!
2. Deep Fakes will teach you Calculus
3. Stable Diffusion 3 is in early access

Google is Racist

TLDR:

After they launched they very much promising Gemini 1.5 last week, and it was rolled out to people, people noticed the image model was racist, it wouldn't generate pictures of white or hispanic people 🤦. Eventually Google apologized and disabled Gemini image generation until they learn how to be more inclusive.

Long Story:

Google fucked up big time this week.

1. They have the least inclusive image generation algorithm in the world
2. They accidentally announced that they will discontinue Gmail!
3. They already killed an AI product.

Last week Google Announced Gemini 1.5 a promising LLM, that uses a unique and new technique, but along it they introduced a new policy inside the algorithm, that was mean to protect results against biased information and output. However they kinda turned the knob to the max, and now Gemini won't generate images of white people.

Now when somebody was asking for an image of the US senators of the 1800s the output was a little unrealistic, as it didn't contain a single white person it it. So naturally people got kinda offended, but not too ofended.

You'd be like, "yeah, fuck those whities dominating everything 🖕".

The major idiocracy is that people asked for images of nazis, and they were all ethnicities but white. That escalated quickly – and the Internet got offended really fast. Because nobody can be a nazi, but a white man, with a short mustache and a very offensive hair style.

Eventually Google had to disable the image generation until they learn how to become a more inclusive company, and stop brainwashing their own employees.

Next Fuck-up by Google

was an email sent to Google's SVP Chris Bakke, which said, please announce that we're sunsetting Gemini team, which Chris read as Gmail. As a result, the 1.5 billion user base of gmail, almost saw an email where google announced that they are killing another of their product. I mean, wouldn't it be fucking ridiculous if it'd happen? Oh we're living on such a fine edge. Loosing my email would be like loosing my entire life, and I trust it completely to one and only single company. Google... oh uh.. I should maybe reconsider my life choices.

Chris Bakke @ChrisJBakke I'm the SVP of Product at Google ($28.8M TC). Last night, Sundar asked me to "fire the entire Gemini team." I misread his message (insanely bad service at my 7br 11bath cabin in Park City) and I accidentally fired the entire Gmail team. Gmail...Gemini. Look at that… https://twitter.com/i/web/status/1760752877218517052 7:46 PM • Feb 22, 2024 606 Retweets 9923 Likes Read 346 replies

​

Fuckup by Google Number 3

was they killed Bard.

Google has a very wide spread tendency to kill products, if there isn't enough traction for them, they don't consider it a viable/interesting product, so they just kill it. For an entire list with a RIP stones see https://killedbygoogle.com/. They try to maintain a decent track of former google products.

Well it's time for Bard to join the rankings, because with the launch of gemini, google replaced entirely bard. RIP Bard 2023 - 2024

​

Massimo @Rainmaker1973 Learning Math from USA Presidents [📹Onlocklearning] 6:38 AM • Feb 5, 2024 119 Retweets 970 Likes Read 37 replies

​

Deepfakes on the rise

TLDR:

Gen-Z are using deep fakes to explain Calculus to remember them easier. Pretty innovative I must say.

Long-ish Story:

Back when I was a student Calculus was probably one of the toughest subject for me, simply put I had a really poor teacher in school and I didn't know shit about math. When I reached uni, I saw Calculus, and was like: "Wtf?! where are the numbers here?". Well GenZ is known for their ingeniosity, and well somebody made use of deep fakes, to create a simple explanation for a differential equations.

I honestly find it genius, and I'd love to learn math from their own authors, using deep fakes.

Imagine learning physics from Isaac Newton, or learning learning how to sing from Taylor Swift? I mean wouldn't that be freaking amazing?

​Stable Diffusion 3, in the early access.​

TLDR:

Stability AI is getting ready to launch their updated model of Stable Diffusion 3.0. Which promises to closes the gap towards midjourney 6, and make this thing closer to reality.

Long Story:

Stability AI, is the only company that provides free of charge models for video and image generation (as long as you can host them on your own hardware), alternatively you can use their API, which pretty democratically priced and allows you to generate images at a very reasonable price.

Well they are finally launching Stable Diffusion 3.0, which means better images accessible for anyone at any price. I'm personally very excited about this, because it means easier access to latest state of the art AI models. They promise up to 8Billion Parameters for the LLM, which means this could generate really high quality images. This means we're pushing the boundaries really far!

It still in early access, and only invited people have access to it, but public release is inevitable.

That's it folks, see you next week ❤️️️️️️️

Happy Prompting!

​

Remember to invite your friends to subscribe at https://​newsletter.texti.app​

This week is all about the big players! They're in the spotlight 🔦
What is it really about? Let's dive in.

The highlights of the week:

1. Texti on Tiktok
2. OpenAI Sora
3. Google Gemini 1.5

Our Newsletter is now on Tiktok!

Our newsletter is now available on tiktok. Follow us to see wrapped content on our page. It's not as rich and enhanced as the email newsletter, but we're slowly getting there. Don't miss-out to stay connect with us, stay in touch through our tiktok profile.

Also follow us on twitter and on Linkedin. We post quite often in the other platforms and this way you'll definitely stay on top of the game.

​Open AI Sora​

TLDR:

OpenAI just showed the capabilities of a new text 2 video tool that can generate videos of up to 60 seconds, and it is incredible, it's 10x better what we had before.

Long Story:

As predicted this year will be all about text2video. We've learned how to make high quality pictures (midjourney is just on another level), now we're expanding our capabilities and pushing our hardware to the next level.

This week, OpenAI presented their fresh out of the oven, tool called Sora, and frankly it's just mind-blowing 🤯. Why? Simply because the quality of the video it's producing is just unbelievable. The shapes, the lines, the motion, the fingers – are still delusional, but other than that it's really good.

It is currently in a closed alpha, where only some influencers have access to it. If you share the word about me, I might be one of them at some point in life 🤞.

Anyways, the tech is still the same, take a text, make a picture, and now generate another (60seconds * 60pictures per second) 3600pictures, put them one after another and voilá you have a video. The thing is though, it's sooo much smoother than what Pika does (check this post for context). It has a story line, it does look like a video, not just a weird cartoon. There are either some clever algorithms behind the scene, either AI has a budget of $7 Trillion, and can afford more computing power – regardless it makes a massive difference.

Based on the numbers alone, we can assume that video generation is going to be expensive and slow. I believe the tech will revert to what we've had about 50 years ago, first we make a 200x200px video check it in low res, to make sure we've got the right result, and then just upscale it like crazy to 4k.
Check the demos from Open AI's website.

Prompt: A beautiful silhouette animation shows a wolf howling at the moon, feeling lonely, until it finds its pack.

Prompt: A litter of golden retriever puppies playing in the snow. Their heads pop out of the snow, covered in.

Well aren't they cute? Awww 🥰.
It still has it's flaws, like physics are often all over the place, it sometimes lacks context, it doesn't really know how to interact with objects, as in if in the generated video, somebody bites a cookie, the bite won't disappear, the cookie will still be brand new. Just take a look at the examples bellow:

Prompt: Five gray wolf pups frolicking and chasing each other around a remote gravel road, surrounded by grass. The pups run and leap, chasing each other, and nipping at each other, playing.

Prompt: Basketball through hoop then explodes.Prompt: Basketball through hoop then explodes.

You can clearly see, that is has problems with physics and the way it interacts with objects. It still tries to copy existing videos into reality.

It is clearly flawed, but that's just the beginning. It's only 1 year old. Just think what's it going to be able to do by the age of 5.

​Gemini 1.5​

TLDR:

Google launched Gemini 1.5, it's extra efficient due it's novel architecture, Transformer + MoE, which roughly means that's an inception of AI-s working on a task, and the killer features is that it's context size is up to 1,000,000 Tokens.

Long Story:

Google made history this week. It made an AI that can understand 1.000.000 Tokens at a time! It also applied a new technological architecture, that is much more clever then before and can use it's own resources way more efficient.

- You ask what the hell this even mean?
- Well, AIs are pretty stupid, believe or not, they have the knowledge of 1000 libraries, but they don't know how to apply that knowledge in a specific situation. So what an AI needs to answer more appropriate is something like a short term memory, where it gathers a bunch of information about the state we are in, and then applies the knowledge it has to provide an answer. Now words for AI are measured in Tokens, because AI has little bit of an intricate way of working, imagine it remembers only syllabus in it's library, where each syllabus is a token.

Now to this day OpenAI's GPT4 Turbo had a max memory of 128k Tokens. This week however this barrier was smashed into pieces, Google's AI is now able to remember up to 1.000.000 Tokens.

- You might why does this even matter? Isn't 128k enough to remember the current state already?
​- Well technically yes, 128k is fine and can handle most of the day to day situation. But if you for example want to add it into a project, or a video, or an audio book, then 128k are not even close to be enough, you need more! (We always need more 😅).
​With 1.000.000 Tokens is now able to remember 1 hour of video, or 11 hours of audio, or 30k lines of code, or even 700k words.
​

- Okay clear, how about that novel architecture what does that mean? What's so revolutionary about it?
​- You see, Google says they used a mixture of Transformer (the most modern algorithm for determining the most appropriate tokens) and MoE (an additional layer which is made from mini-AIs expert in a specific field). So basically a way to orchestrate a lot of information in a very smart way. These technical novelties make this a new industry top. This makes it stand-out of the crowd.

- Alright gotcha, it now makes a little more sense. Thanks.

The 1Million tokens is a really big deal, it means it now can go way further down the road. It can now actually build a big project, not of an enterprise level, but for most small businesses this is going to be way more than enough. 30k lines of code isn't something incredible in terms of amounts, but it means, the AI can handle a lot more in it's short term memory. Now combining Sora and the capabilities of Gemini 1.5, we could in theory generate a 1 hour long video, which would contain and be able to figure out an entire story. Honestly sounds fascinating and crazy.

Tomorrow's Christopher Nolan might be any kid with access to internet.

That's it folks, see you next week ❤️️️️️️️

Happy Prompting!

​

Remember to invite your friends to subscribe at https://​newsletter.texti.app​