OpenAI's Random GPT Name Generator Room

Google dropping $32B on Wiz, when they could just vibe coded a clone in an afternoon... ¯\_(ツ)_/¯

How many are finally paying the technical debt using genAI tools to automate & fixing all those boring tasks in the backlog that kept getting ignored?

Never did Studio Ghibli anticipate how many gigawatt-hours of electricity would be consumed to generate Ghibli-style memes.

Kevin Scott, Microsoft's CTO - thoughts on DeepSeek and Chinese AI companies.

Kevin Scott, Microsoft's CTO, his thoughts on how AI code is following the trend of making it easier for people to talk to machines. You will still need good programmers to build and fix complex stuff...

Some Llama.cpp code is being written by AI. This is not your typical python hello world app. This is low level C/C++

Long but very good read on the recent events in AI: Nvidia Moat, DeepSeek, LLaMa, OpenAI, Anthropic etc

Wow :) did I read this correctly? "DeepSeek’s singular focus on research makes it a dangerous competitor because it is willing to share its breakthroughs rather than protect them for commercial gains."

Turns out, shipping a solid model is better than having your marketing team blast AI hype headlines on all cylinders. Less talk, more shipping impressive models! DeepSeek is now nr 2 free apps. Let’s see how will OpenAI pricing evolve from here…

For a nation of 1.4B people talent isn't a resource that's lacking. They've invested big-time in education, hard sciences, incredible manufacturing capabilities. Why is this a surprise?

Cheers to the OpenAI engineers being oncall right now.

Books published before ChatGPT, I see them differently now. They’re basically backups of organic intelligence... human knowledge, creativity, hallucinations and all.

Don’t use LLMs as your primary doctor, financial advisor, or lawyer - but as a second option - hell yeah!

Today I learned that there is a wiki page dedicated to AI doom predictions. Yann LeCun leads the chart with the P(doom) value of <0.01% Bottom of the list: Eliezer Yudkowsky & Roman Yampolskiy.

Quick update on ChatGPT's impressive growth since its release 2 years ago: - Hitting 300 million active weekly users - Over 1 billion messages sent to chatGPT, daily - 1.3 million developers in the US, with an expanding international user base

Amazon’s new LLMs now take the crown for the “cheapest price for intelligence” that Google coined.

You might be surprised by how many people tell me they use LLMs as their personal therapists. This has proven to be a valuable use case, especially for people from cultures where seeking therapy is taboo or stigmatized.​​​​​​​​​​​​​​​​

Why would a company that says it's close to achieving AGI need a Chief Marketing Officer? Wrong answers only.

It's great that LLMs are making coding more accessible, while also helping people understand [sometimes the hard way] the complexity of software engineering. Building production-ready systems requires much more than just typing code.

From the Amazon Nova Family of Models technical report: Just a reminder of how freaking fast Gemini models are. They might not be the smartest, but they're super fast! Also, strange choice to compare with the older 001 models when the 002 have been out for some time now.

Amazon reminded everyone that they're in the game too. I'm excited for early 2025: Claude Opus 3.5, Gemini 2, Grok 3, OpenAI full o1, LLAMA 4 [maybe], and seeing Chinese labs pushing open-source models.

Amazon just dropped a gazillion videos from re:Invent 2024. If only there was a way to summarize them... in 1 word.

I used Microsoft Copilot for help with a Cosmos DB query, failed multiple times. It didn't solve the problem. Something tells me that big companies can solve a lot of real world problems by shipping tools that give accurate answers to their own products. Looking at you Microsoft, Google & Amazon!

Good stuff! 5-Day Gen AI Intensive Course with Google www.kaggle.com/learn-guide/...

One practical idea I consistently use is keeping a few browser tabs open with different models from OpenAI, Anthropic, Google, and Meta. I test them daily with various problems coming my way, this helped me quickly understand their strengths and limitations.

I can see a future where powerful AI models run securely on your phone, huge memory, that truly understands you. It would know your social graph / friends, emails, notes, photos, the apps you use… This personal “genie in a bottle” will help you get things done faster and more efficiently.

Tristan Harris: "What I don't think what we're prepared for is we're going to live very soon in a world where 10 years of scientific research could happen in like a single month." I thought these guys were serious, but this comment easily wins the Bullshit of the Year award.

I’m not sure how useful it is to automatically store this in ChatGPT’s memory 

Fun fact: Horses originally evolved in North America, migrated to Asia, and became isolated there during a warm interglacial period. They eventually died out in North America, only to be reintroduced by the Spaniards in the 1500s.

I really liked this part when I was at Meta. Code [data] wins arguments.

OpenAI thinks I talk too much.

I've renamed all my buggy bash scripts. I now call them AI Agents! These bad boys nail about 85% of tasks I throw their way, crushing the performance of standard LLM-powered [AI Agents] alternatives. Thinking of raising $35 million for them. Hit me up in the DMs if you wanna invest.

OpenAI compliance with GPDR [I guess] is back firing

Geoffrey Hinton: we actively learn for about 10^9 seconds or ~32 years. Our brain has 10^14 connections, which makes that we have a ratio of 100,000 connections in our brain for every second we experience in the first 32 years of our life.

Elon Musk, early life lessons

Logan Kilpatrick's podcast - quite under rated (152 views so far) ! His episode with @simonwillison.net is packed with good stuff youtu.be/rLcKbvmegag

Fort Ross, California: Prayers backed up by artillery.

Everything takes longer than you think.

Every solution breeds new problems.

The tech world experienced a seismic shift two years ago today. It sparked a level of excitement in me that I haven't felt since the early days of programming This is the reason why combined market cap of companies like Nvidia, Microsoft, Meta, and Google are now worth trillions more:

This is like Formula 1, but for software engineers.

What is intelligence?

Arab merchants, for centuries, were telling tales of the dangers of collecting spices from mythical beasts. The point was to jack up the price and keep the Greeks and Romans clueless. You can see parallels with AI labs that tried to make LLMs sound all mysterious, and AGI's just around the corner

Ukrainian soldiers… carrying their brother… They’ve seen it all in these > 1000 horrific days of war.

The democratization of software development is a major win of the current genAI wave. It’s a true unlock for millions of builders around the globe. This will get us in an era of abundant / cheap software that solves more problems we didn’t had the bandwidth to solve.

The theme I see in this wave of genAI: human augmentation and, on occasion, human replacement. Stay curious, experiment, and learn by doing - using AI tools. You will be less competitive if you don’t.

I’m more concerned with the growing trend of teenagers / young people forming emotional bonds with AI characters and imaginary companions, than the hypothetical emergence of a self-replicating artificial superintelligence. Social media echo chambers are a joke in comparison.

AI spending surged to $13.8 billion this year, more than 6x the $2.3 billion spent in 2023—a clear signal that enterprises are shifting from experimentation to execution, embedding AI at the core of their business strategies. Code copilots lead the charge with 51% adoption!

Good tips for searching stuff on the Internet: gwern.net/search