Rob Leclerc

Just posted an article, a General Theory of Neural Networks -- an idea that I've been thinking about since my PhD. robleclerc.substack.com/p/general-theo…

@cryptopunk7213 an agentic model that is training itself will likely be better at agentic work compared to chatbot training.

fuck me china just launched the 1st AI model that autonomously built itself... and its as good as claude opus 4.6 and gpt-5.4 - minimax M2.7 trained itself through 100+ rounds of autonomous self-improvement. 30% gain. No humans involved - what the actual f*ck - model now handles 30-50% of the AI lab's OWN AI research - beats gemini 3.1 at coding and pretty much matches opus 4.6 + gpt 5.4 😶 (china used to lag now they match - doesn't require crazy hardware to run (single a30 gpu) - absolutely CRUSHES tasks: financial modelling, coding, openclaw - one-shotted the chinese have officially caught up. self-improving ai is a real thing. all researchers did was set an objective and the model figured the rest out. i wasn't expecting this from minimax. im now wondering wtf deepseek is going to be like.

the problem with skills is that the ai often does’t use them.

does agentic engineering mean coding with agents, or coding an agentic system? that distinction is not really clear yet.

@iruletheworldmo to be fair, people have been talking about creating openclaw roles for two months, then gary comes along and passes it off like he invented it by naming it gstack…

so this guy has lost his sense of meaning and has been oneshotted by claude but he's going after people using claude and sharing their ideas because they've accepted their role in this new world and he can not. shall be an interesting year for sure.

building software right now is like playing a video game. the loop between thought and working product is so tight, and the leverage you get from small moves is crazy

i can set up twilio & elevenlabs and have my openclaw agent call me, all in under five minutes, but i can’t get google oauth to work after 30 minutes.

took spiraling and pinealon for a month. had ZERO impact on my 8-sleep sleep score. buyer beware.

this could be the beginning of a taiwan invasion or it could be posturing to get the u.s. to back off iran.

i used to think i liked coding, but now when i have to deal with a sticky bug in code, i just get annoyed. apparently I just like to build and code was the only tool i had to do it.

@FutureFetish007 we’ll have to agree to disagree. i think you’re missing the big picture and that your view is colored by an obvious dislike of trump.

@robleclerc And I hear reports we are taking weapons out of S Korea etc Yeah this is a clusterf**k of bad planning. Iran is no Venezuela. China could lean on Taiwan hardcore. Meeting in 2 weeks with Xi

chinese equipment failed badly in Venezuela and iran, but if they wanted to take taiwan, this would be the time to do it with the u.s. tied up in iran.

@rieglobe @AlphaSchoolATX sad to say, but in 99% of cases those accepted will be wealthy kids from hard core tiger parents who gave their kids a rich cv from tutors, home schooling, top private school, travel, and other extracurricular activities.

What if investors funded @AlphaSchoolATX tuition for kids who can’t afford it in exchange for a small share of the $1M they’re expected to make by graduation? Tuition is fully refunded by the school if they don’t make it. Instead of student debt, students graduate with financial confidence and entrepreneurship through Alpha. True win win. This is the way.

@r0ck3t23 this is the way

Perplexity CEO Aravind Srinivas just shattered the greatest illusion of the AI arms race. The entire market is waiting for a single, god-like superintelligence to win the entire board. The physics of compute are forcing the exact opposite outcome. Models are not converging into a single monopoly. They’re violently fracturing into hyper-specialized execution nodes. Srinivas: “Towards the end of 2025, what happened was models started specializing. Even within coding, which you think might be a specialization, OpenAI’s Codex models and Anthropic’s Claude models are very different in terms of what they’re good at.” Bet your entire enterprise architecture on a single AI provider? You’re hardcoding your own ceiling. You don’t want a generalized model that’s “okay” at everything. You want a swarm of apex specialists. One ruthlessly optimized for syntax. One for visual synthesis. One for predictive reasoning. The future is not one AI. It’s the instantaneous orchestration of the absolute best compute for the exact task at hand. Platform lock-in is suicide. Srinivas: “Enterprise users are always selecting multiple different models all the time. That’s actually one of the value propositions of the Perplexity product. You don’t have to feel locked into one model provider, you don’t have to have one horse in the race.” Traditional tech giants are desperately trying to trap users inside their specific algorithmic ecosystem. Winning operators completely bypass the vendor war by becoming model-agnostic. When the foundational intelligence of the world is leapfrogging itself every three months, brand loyalty is a massive liability. The operators winning the next decade won’t care whether OpenAI, Anthropic, or Google trained the model. They’ll plug into an agnostic orchestration layer that autonomously routes to whichever specialized network currently dominates that exact sector of the board. The highest-leverage position is no longer building the intelligence. It’s directing the orchestra. Srinivas: “This is one particular skill, writing is another skill, being good at images and videos is another skill. You can hope that Perplexity figures out which model is best for what purpose, and you just have to come to the product and use it.” Multi-trillion-dollar hyperscalers burning billions fighting the model wars. Sovereign orchestrator bypasses the entire war. Harvests the output of all of them. You don’t need to be an expert in the underlying architecture of a dozen different foundation models. You just need to command the routing engine. When AI transitions from a monolithic product into a fractured grid of specialized utility nodes, the ultimate monopoly belongs to the orchestrator that abstracts the complexity. Foundation model builders became interchangeable plumbing.

@noahzweben where did that first overlay window come from? i have all the latest versions installed on a max plan and have been unable to get this work. i even tried having cc debug it.

Remote Control - Session Spawning: Run claude remote-control and then spawn a NEW local session in the mobile app. * Out to Max, Team, and Enterprise (>=2.1.74) *Have GH set up on mobile (relaxing soon) * Working on speeding up session start-time

@DaveShapi i’ve spent the last month with openclaw and it may be the economics and reliability of large context windows may force a greater division of labor—at least in the short term. and by agent i mean ai+harness.

More like in 2 years and it will be thousands of ephemeral agents plus a few dozen permanent agents

in 10 years we’ll have 10x-100x agents for every human. billions of them may need human services every day.

BREAKING: New startup "RentAHuman" allows AI agents to rent humans to perform tasks they cannot physically perform themselves.

drag-and-drop airstrikes.

@PalantirTech vibe combat

"This is Maven Smart System—Palantir’s software as a service product that we are deploying across the entire department."

“I care about one thing and one thing only: that the 18, 19, 20, year old kid—who had no choice in where he went, or what threat he was facing—I want him to win and come home. That’s why we do it. Palantir is very helpful in delivering this.” Cameron Stanley, Chief Digital and Artificial Intelligence Officer of the Department of War, shares how the DoW is driving enterprise-wide adoption of data, analytics, and AI to generate decision advantage—and what it takes to move cutting-edge technology from the lab to the warfighter at speed.

agentic engineering isn’t writing a prompt, reviewing the output, and repeating. it’s carefully reading through the thinking traces to understand the bugs and design choices so that you can better steer the model.