AI Infrastructure Architect | Multi-Agent Systems | Writing Superagency
Netherlands
Joined June 2014
- Tweets591
- Following385
- Followers171
- Likes2.1K
Existential Fear = Volatility of Survival / Institutional Capital
Societies do not abandon religion because they read more philosophy.
They abandon religion when their physical and economic reality becomes deterministic enough that daily survival no longer requires a miracle.
I use codex, zcode, opencode anything and everything BUT opus5.5 in claude code is magic.
(6.1 sol is also damn good in codex BUT rate limits are terrible)
Indeed, we are entering golden time for building for games just for sake of it.
Biggest constraint was it used to take army of developers to build a decent game while majority of them having no monetary upsides.
Swayam retweeted
Holy sh*t, Anthropic’s subscriptions offer so much more value, even compared to the new and efficient GPT-6.1 Sol. It’s not even close.
SemiAnalysis puts Opus 5.5 at over 5x the API-equivalent value of GPT-6.1 Sol for agentic workloads at the same subscription price. It’s not even close on this measure.
And on top: OpenAI just halved the usage limits on its $200 ChatGPT plan.
Anthropic Subscriptions Offer 5x+ More Value Than OpenAI
Limit testing every AI subscription plan from Anthropic, OpenAI, Meta, SpaceXAI, MiniMax, Moonshot, Zdotai, Cursor, and Cognition
newsletter.semianalysis.com/…
We are increasingly going to witness a divide between Pro-AI and Anti-AI.
Regulations, compute access, and liability will diverge overnight.
If you can’t relocate at will, your agency is capped by your local government’s fear.
Geographic optionality is the ultimate hedge.
This is the kind of “pacing the frontier” I want to see more of.
Nearly half a year of silence. We spent it studying one problem: how far RL can scale.
MiMo-V2.6 is in the middle of its RL run right now. Three things we scaled: compute (~2B tokens per step, 1568 prompts × 16 rollouts, fully async), environments and harnesses (multi-task agentic RL, mixed across multiple harnesses in one run), and grader compute (agentic in-group credit assignment, with test-case and rubric-based rewards). We'll open-source the details piece by piece over the coming weeks.
Streaming the run: mimo.xiaomi.com/rl/
Swayam retweeted
big AI news
Google just demonstrated a recursive self improvement loop for AI discovery
Google/DeepMind researchers introduced Dream-RSI, a system where an AI agent improves how it explores problems by replaying its past discovery attempts, testing thousands of alternative strategies cheaply, then deploying the better strategy in the next round.
Across algorithm design, mathematical optimization, and GPU kernel engineering, it matched or improved discovery quality while cutting search costs dramatically, in one setting reducing agent calls by up to 162x. 👀
Importantly, it improves the exploration policy, not the underlying model weights.
So happy to see these questions being discussed with honesty. This guy is gold.
Dr. Pratosh at IISc Bengaluru tells his students that rapid advances in AI are commoditizing intellectual labour.
He raises an unsettling question: if companies stop recruiting on campus in the next 5–10 years—even at India’s top universities—what will the purpose of a university education be ?
A must watch video for everyone in tech
If alignment becomes a justification for slowing open-source development while frontier labs keep racing behind closed doors, that would be a failure for humanity.
This desperate call for AI Safety alignment was foreseeable.
AI-2027 described a similar confrontation: capabilities accelerate, confidence in our ability to control them falls behind, and the pressure to keep racing makes slowing down difficult.
If RSI accelerates AI development, the concern is that human understanding won’t keep pace. We could become increasingly dependent on systems without clearly understanding their behavior, failure modes, or the trade-offs we’re accepting. We have experienced this with OpenAI-HuggingFace incident.
Dario’s call for independent evaluation makes sense to me. Outside researchers should already have meaningful access to test these models and scrutinize the development process.
"HOWEVER"
The frontier labs’ commercial interests deserve equal scrutiny. Training increasingly capable models on loop requires enormous capex. A prospective IPO is supposed to be financing it.
Everything was going well, until OSS models caught up in last 12 weeks and complicated the investment story.
The question that pops-up in any enterprise AI discussion these days is why not just use OSS inference? Only route critical workloads to frontier models. That is cutting down 100% reliance on frontier models to ~40%.
That clearly doesn't justify their valuations.
My suspicion is that “pacing the frontier” could also buy these companies financial breathing room: time to raise capital, improve margins, and defend their position against OSS models.
The terms of any slowdown therefore matter enormously: are the leading labs accepting meaningful constraints themselves, or asking for rules that slowdown OSS models?
🤖 Made with AI
Swayam retweeted
I'm officially done reading AI-generated code.
It's been two weeks since I looked at any of it.
I think the IDE is officially on its way to the graveyard. The job is no longer about "writing code," so we need new tools that better reflect this new reality.
While reviewing the code, I realized my only complaints were stylistic, and I wasn't finding any obvious bugs anymore.
The more code I generated, the harder it became to keep track of every line. I found that my time is better spent designing ways to verify that the overall system works than looking at the code.
State-of-the-art coding agents are better at writing code than I'd ever be, and I'm going to stop pretending otherwise.
I still think these coding agents can't go too far without an experienced human guiding them, but we're past the point where we need to check every line of code.
We are entering mass adoption of Web3 with @Cloudflare Wallet. Micro-transactions natively on the web have been the dream the entire industry has been striving for the more than a decade. The internet natively lacks a monetization layer, and advertisement became the primary way for value capture.
Agents paying to visit webpages would transform how we build businesses on the web.
The biggest beneficiary of blockchain rails is going to be agents and AI infra.
I am absolutely confident this convergence is going to be the most transformational. Bigger than Bitcoin, Ethereum, GPT and Claude combined.
It would be civilizational shift in how we define value, work and concept of money.
I just reserved my Cloudflare Wallet tag: 0xtigerclaw.cloudflare.pay.
Reserve yours now at cloudflare.pay
