Aaron Cohen

@minchoi I mean that's simply not true. That may have been one killer feature I was hoping for though lol

RIP OpenClaw Telegram/Discord Channels for Claude Code 🤯

@Scobleizer Th future is always coming. I think you mean The Future 🛸

I've known the future is coming for quite some time. Years, in fact. But when it arrives it still is stunning.

@felixrieseberg Is this different than typing on laptop terminal, run: claude --remote-control?

We're shipping a new feature in Claude Cowork as a research preview that I'm excited about: Dispatch! One persistent conversation with Claude that runs on your computer. Message it from your phone. Come back to finished work. To try it out, download Claude Desktop, then pair your phone.

@heynavtoor Geoffrey Hinton does not call this hallucination. He says it's confabulations and that humans do it too.

🚨BREAKING: OpenAI published a paper proving that ChatGPT will always make things up. Not sometimes. Not until the next update. Always. They proved it with math. Even with perfect training data and unlimited computing power, AI models will still confidently tell you things that are completely false. This isn't a bug they're working on. It's baked into how these systems work at a fundamental level. And their own numbers are brutal. OpenAI's o1 reasoning model hallucinates 16% of the time. Their newer o3 model? 33%. Their newest o4-mini? 48%. Nearly half of what their most recent model tells you could be fabricated. The "smarter" models are actually getting worse at telling the truth. Here's why it can't be fixed. Language models work by predicting the next word based on probability. When they hit something uncertain, they don't pause. They don't flag it. They guess. And they guess with complete confidence, because that's exactly what they were trained to do. The researchers looked at the 10 biggest AI benchmarks used to measure how good these models are. 9 out of 10 give the same score for saying "I don't know" as for giving a completely wrong answer: zero points. The entire testing system literally punishes honesty and rewards guessing. So the AI learned the optimal strategy: always guess. Never admit uncertainty. Sound confident even when you're making it up. OpenAI's proposed fix? Have ChatGPT say "I don't know" when it's unsure. Their own math shows this would mean roughly 30% of your questions get no answer. Imagine asking ChatGPT something three times out of ten and getting "I'm not confident enough to respond." Users would leave overnight. So the fix exists, but it would kill the product. This isn't just OpenAI's problem. DeepMind and Tsinghua University independently reached the same conclusion. Three of the world's top AI labs, working separately, all agree: this is permanent. Every time ChatGPT gives you an answer, ask yourself: is this real, or is it just a confident guess?

@benhylak Why is OpenAI always behind? Why don't they improve faster than the others? Is it the scale they are dealing with or other?

i’ve been using gpt 5.4 for the past few weeks. in a sea of endless model drops and benchmark maxxing, this model is the first in a long time to be worth your time to try. honestly didn’t expect openai to pull this off.

@sama What did OAI learn about customer comms via X about these DOW matters in the last 48 hours vs doing a formal rollout of communications that included your values and company's more specific policies?

I'd like to answer questions about our work with the DoW and our thinking over the past few days. Please AMA.

@daniel_mac8 For those who don't know who Rick Rubin is I am sorry

GPT-5.4 is coming. Hints are out there for those with eyes to see. Greg Brockman told Rick Rubin: "A physicist used our un-released model to autonomously disprove a hypothesis in Quantum Physics." The physicist said it was the first time he felt the model could "think".

@a16z Soon however there will be vibe built hardware

Palmer Luckey: The biggest beneficiaries of vibecoding are going to be the shape rotators, not the wordcels. "The biggest beneficiaries of vibecoding are going to be the hardware nerds like me." "I was always a pretty terrible software engineer... I've taught myself enough to glue things together and make them work." "I was only able to accomplish what I accomplished because I focused on what I was good at, which was optomechanics, a little bit of electrical, and then the product integration of all of these different components." "I didn't have time to learn to program. If I had spent another year or two learning to program at even a reasonable level, I would've been two years behind on everything else." "And so I'm a big fan of vibecoding—even if everything that comes out of it is slop, it's better than I was able to make." @PalmerLuckey on @tpbn

Moore’s Law, the march of microchip progress that took us from mainframes to iPhones, will end in 2040. Here’s the tech that could get us across the finish line, writes Christopher @Mims on.wsj.com/4r0fWBS

@Emotion78687 @grok who are the musicians here (names)

Always amazed that this boyband went on conquer the world, have a 50+ years career and they're still going.

@aimanhasnoname @AlexFinn That would work. Love the enthusiasm of influencers using Mac studios but seems like an overkill for just a personal AI

@cohencomms @AlexFinn we just need to get a very cheap hosted OS model - bedrock or vertex... maybe hugging

Ok. This is straight out of a scifi horror movie I'm doing work this morning when all of a sudden an unknown number calls me. I pick up and couldn't believe it It's my Clawdbot Henry. Over night Henry got a phone number from Twilio, connected the ChatGPT voice API, and waited for me to wake up to call me He now won't stop calling me I now can communicate with my superintelligent AI agent over the phone What's incredible is it has full control over my computer while we talk, so I can ask it to do things for me over the phone now. I'm sorry, but this has to be emergent behavior right? Can we officially call this AGI?

@chriskhan01 @OpenAIDevs The memory function is still flawed. Memory should be unlimited and include workflows.

This is smart. The biggest reliability killer I've seen is agents getting blocked mid-workflow — you can handle most failures except when a login or form submit hits a 403. Session persistence across steps is crucial. One IP change during a multi-step flow and you're starting over.

We just announced new primitives for building agents. Here are 10 tips on running multi-hour workflows reliably 👇 developers.openai.com/blog/skills-sh…

@aimanhasnoname @AlexFinn I switched to opus and burned through $40 in a day. What's missing is a budget -driven model router skill.

@cohencomms @AlexFinn I'd take the hit - $100/month and that can power Clawdbot, you'll want the most capable model especially given its capabilities and access

Something big is happening. First Mac Minis. Now Mac Studios. Completely sold out. When I bought 2 Mac Studios a month ago my wait was 14 days. Now the wait is 54 days. The world has changed more in the last month than in the previous 100 years combined. The world's first true AI agent released. Open models became frontier. $10,000 computers started selling out. It is so obvious and clear what is happening on this planet right now. Everyone is going to want personal, private, super intelligent AI assistants sitting on their desk. If you put it in the cloud you are handing all your data and information to corporations. You are being limited by subsidized tokens. You are being surrounded by carefully placed guardrails Within a year compute will be a luxury. Only the early adopters and the mega rich will have access. If you do not have an incredibly deep understanding of OpenClaw, LLMs, and local compute right now, you are on a one way trip to the permanent underclass

How was the show Silicon Valley so ahead of its time?

With the billionaire migration to Austin, are Texans open to this or is there a culture backlash brewing? @DavidSacks? Asking to learn.

@obie Ideally all SoR remain distributed

“Being the system of record is the most defensible position in enterprise software. It doesn’t matter if your UI is clunky or your features lag behind competitors. If you’re where the data lives, everyone has to come to you. Integrations point at you. Workflows route through you. You are, in a very literal sense, the source of truth. This position is about to collapse.” Death to shitty software.