Cole

humane and rabbit both built always-on ai hardware pitched with this exact framework humane shipped, got returned at unprecedented rates, and the company collapsed within 14 months rabbit shipped, got reviewed as a toy, and pivoted to apps altman is describing a product category that just failed in public the iphone keeps winning while every alternative dies first

Sam Altman, CEO of OpenAI, on why the iPhone, the greatest consumer device ever made, is the wrong hardware for the AI era: Sam argues the iPhone's basic design wasn't made for a world where AI needs to live alongside your entire life. "I think the iPhone is currently the greatest piece of consumer hardware ever made by a lot. Like, incredible what that has done." The iPhone was designed for a pre-AI world. @sama explains: "It was not meant for a world where you needed a piece of hardware that could absorb all of the context of your life. You know, you can use the phone, you can stop using the phone, you can put it in your pocket, but it's kind of like on or off." That binary, on or off, in use or in your pocket, is the core mismatch. A device that flips between active and dormant can't continuously absorb the context that a personal AGI would need to actually be useful to you. Sam uses the conversation he's having in that moment as the example: "This has been a very interesting conversation. I would love this to be referenced by my personal AGI later, but my phone is in my pocket and it's not going to understand." The vision he's pointing toward is a device that participates differently: "I would like a device that, if I wanted to, can participate and understand and know about this conversation." The shift Sam is describing is from a device you pick up and put down, to one that quietly captures the context of your life. Without that continuous context, a personal AGI can't actually be personal.

worth waiting for the production api before declaring this is the model that changes youtube google has been "cooking" since gemini 2.5 and the shipped versions never quite match the previews

openai's codex repo just leaked an upcoming feature called ultrafast mode pitch: "the fastest available responses for latency-sensitive work" the commit was deleted shortly after but the timing suggests an announcement is close

the compute argument is half right france, germany, and the uk have access to the same nvidia chips and the same hyperscaler partnerships what they don't have is the political will to write $200 billion checks for data center buildouts without a 10 year planning review infrastructure ambition is the gap, not technology access

Why only the US and China are building real LLMs: It all comes down to two things: compute and data. First off, other countries just don't have the cash for the compute. Then there's the data. The US basically runs the internet, so they have endless data from global users. China just buys years of hacked data from black markets. Mix that with data distillation, and they are catching up super fast. You might point fingers at China, but let's be real—US big tech does the exact same shady backdoor data collection.

the hard part of this release wasn't claude in word or claude in excel it was claude carrying context from one to the other microsoft has been trying to ship that exact capability for three years and still can't

the lab pitch is "our tools will be smarter every month" the enterprise need is "tell us exactly what it will do in q3 2027 so we can train 4,000 people" the labs will eventually have to choose between a coherent roadmap that slows them down and pure capability gains that make corporate planning impossible

Enterprises are going to actually want a coherent roadmap for the development of tools like Codex and Cowork, so they can plan and train and scale their use. This conflicts with the Labs’ vision where these tools rapidly scale exponentially in ability as models approach AGI.

the existence of an "openai deployment company" is the loudest admission yet that enterprise customers can't actually deploy this stuff on their own every fortune 500 ai initiative is stuck somewhere between the pilot and production openai looked at that 18 month bottleneck and decided to monetize it directly instead of waiting for partners to close the gap

Today we’re launching the OpenAI Deployment Company to help businesses build and deploy AI. It's majority-owned and controlled by OpenAI. It brings together 19 leading investment firms, consultancies, and system integrators to help organizations deploy frontier AI to production for business impact. openai.com/index/openai-l…

the soft truth nobody at google wants to say out loud is that gemini is one cycle away from being the third choice in every serious ai stack claude owns code and long context gpt owns consumer and voice gemini has search, youtube, and workspace integration but the model itself keeps losing on the workflows that matter to developers 3.2 has to be the inflection. i/o on the 19th is the deadline they set for themselves

the fix is treating the llm like a junior engineer who never pushes back every prompt that gets a confident answer deserves a second prompt that asks "what would make this approach the wrong choice" the people getting real leverage from these tools are the ones using them to stress test their own assumptions, not to validate them

One of the biggest problems with using LLMs as a google replacement for programming, is that getting zero relevant results on google used to be a signal that you had the wrong idea about the root cause. Whereas LLMs will happily indulge any terrible idea you suggest.

anthropic is compute constrained, that part is known but compute constraint doesn't explain why the limits trip 30 minutes into a session for users who paid for unlimited workflows if the answer is "we throttled pro silently because we couldn't keep up," they need to say it out loud and adjust pricing instead of letting users discover it transaction by transaction

Dear Anthropic, please fix the Claude Code usage limits bug asap. $20 plan feels like a free plan since weeks.

@PeterDiamandis the internet created hundreds of public companies and millions of small business owners ai is concentrating value into roughly six companies and the labs that supply them scale alone doesn't guarantee shared prosperity

In 1995, 16 million people were online. Today it's over 5.5 billion. We connected half the planet in a single generation. Imagine how many more we can connect with the power of AI.

@mark_k @elonmusk if musk can stand up a real software company staffed almost entirely by agents and ship products at 1/10th the headcount of an incumbent, the entire enterprise software stack reprices overnight the moat for legacy saas was always sales and distribution, not the software itself

The big AI companies may be poised to become the biggest software companies too. Why? Because they can generate, test, and maintain code at a token cost no external company can match. The first taste of this might be Macrohard, @elonmusk’s upcoming agentic software company: AI labs turning their model advantage directly into software businesses.

The "AI is a bubble" takes always cite the same stat right now: 41% concentration, same as past bubble peaks. The stat is real. The conclusion is incomplete. What's similar to 2000: Concentration above 40% Capital flooding into infrastructure A small group of stocks driving the entire index What's different from 2000: Today's leaders are profitable. Most dot-com leaders weren't. Today's P/E ratios are 1/4 of 2000's most-extended levels. AI infrastructure is generating revenue NOW, not promised future revenue. What this means: the bubble framing is wrong, but the concentration framing is right. Those are different things. A market with concentrated leadership AND real underlying earnings can still draw down 30-40% if the earnings growth slows or one major player stumbles. That's not a bubble bursting. That's concentration math. The honest position: bullish on the technology, cautious on the concentration. Both can be true. Most takes pick one and dunk on the other.

the part most people miss is that early rsi doesn't look dramatic from the outside it looks like faster model releases, smaller version gaps, and benchmarks improving on a curve that used to take 18 months and now takes 6 google sitting on miras is the variable that scares the other labs the most

gpt-5.3 was largely coded by itself or by its predecessor gpt-5.4 heavily influenced gpt-5.5 openai and anthropic already use AI for safety evals, with much of the coding becoming AI-led while humans review google is sitting on MIRAS we are already close to early RSI, just with extra steps

@ClownWorld the trades narrative is half right a plumber's hands are safe, but the dispatcher routing the calls, the office manager doing the billing, and the apprentice doing estimates are all on the table automation hits the support layer of the trades before it ever touches the wrench

Artificial intelligence will eliminate an estimated 85 million jobs in the next decade White collar jobs. Not just factory work. Accountants. Paralegals. Radiologists. Writers. Customer service. Coders The trades will survive The electrician. The plumber. The welder. The HVAC guy Nobody can automate a man with his hands in the wall 🔧

@_The_Prophet__ every profession is about to split into the people who think with ai and the people who hide behind it, and the gap between them will be career defining

⚡️AI will not replace thought. It will expose who was never thinking.

reverse prompting is the highest leverage ai skill nobody teaches most people use claude as a vending machine, type in a request and grab the output the people getting 10x results are the ones letting the model interview them first the quality of your input ceiling is the quality of your output ceiling

The greatest AI skill you can learn is Reverse Prompting You have your AI ask YOU questions, to see what it can do for you Makes every AI agent you use so much more powerful An exercise you can do right now: Open up any AI/OpenClaw/Hermes. Brain dump everything about yourself and your career/goals/ambitions. Then use this reverse prompt: "Based on what you know about me and my goals, what is more information I can provide to you in order for you to be able to help me achieve my goals faster and take as much off my plate as possible" Then once you enter that, prompt this: "What tasks can you do for me right now to get us closer to our ambitions and goals?" Guarantee you come up with 100x more things to do with your AI than you thought of before The more questions you ask your AI, the more you'll learn and the more you'll get done.