Sabitlenmiş Tweet
syntax.oleacomputer.com
Beta now available
Send feedback to hello@oleacomputer.com
English
cameron
2.3K posts
cameron
@camweb3_
building https://t.co/nULPz9jEcD | prev Devrel @Coredao_org, @HederaFndn
Michigan, USA Katılım Haziran 2018
3.4K Takip Edilen787 Takipçiler
cameron retweetledi
JUST IN: OpenAI raises $122,000,000,000 at $852 billion valuation.
English
@adamdotdev and look at who has actually shipped product before lol
garry is such a fraud
English
CORETOSHIS and $CORE holders where do you project $CORE to reach by the end of the year 🙌?
@grok you can join the chat and help us predict, there's $CORE for you to win🙂
English
@DavidKPiano The virality of this is like a planted seed to see the effects of anonymous distributed reviewers. Anthropic making the assumption the audience will feed back into the LLMs anyways
English
Ironically this is probably the first time that actual humans are carefully & thoroughly reviewing the Claude Code codebase
Chaofan Shou@Fried_rice
Claude code source code has been leaked via a map file in their npm registry! Code: …a8527898604c1bbb12468b1581d95e.r2.dev/src.zip
English
cameron retweetledi
NEW research from CMU.
(bookmark this one)
The biggest unlock in coding agents is understanding strategies for how to run them asynchronously.
Simply giving a single agent more iterations helps, but does not scale well.
And multi-agent research shows that coordination > compute.
A new paper from CMU proves this with a practical multi-agent system.
CAID (Centralized Asynchronous Isolated Delegation) borrows proven human SWE practices: a manager builds a dependency graph, delegates tasks to engineer agents who work in isolated git worktrees, execute concurrently, self-verify with tests, and integrate via git merge.
CAID improves accuracy over single-agent baselines by 26.7% absolute on paper reproduction tasks (PaperBench) and 14.3% on the Python library
development tasks (Commit0).
The key insight is that isolation plus explicit integration beats both single-agent scaling and naive multi-agent approaches.
For long-horizon software engineering tasks, multi-agent coordination using git-native primitives should be the default strategy, not a fallback.
Paper: arxiv.org/abs/2603.21489
Learn to build effective AI agents in our academy: academy.dair.ai
English
cameron retweetledi
@howdymerry @thattallguy syntax.oleacomputer.com uses autoresearch. Will test with this repo and see how it improves ‘research’ section and portfolio management
English
I took Karpathy's Autoresearch concept and adapted it into AutoPredict: a research framework for evaluating, backtesting, and iteratively improving prediction market trading agents
AutoPredict evaluates agents on
- forecast quality
- calibration
- execution (slippage, liquidity, and fills)
- drawdown and risk adjusted returns
It also supports domain specialists for weather, finance, and politics under a shared evaluation harness
This framework is NOT for building agents but for agent improvement via a evaluation + mutation + selection loop
English
@aidenybai Probably have an extreme fear of death because you don’t have a belief in god
English
@MatthewRideout @artyomvnsv Skill issue. Create infrastructure, environments that recursively check against your constraints in limited design files. And ensure version control is tight as you iterate. If the quality is degrading this is once again a problem with the trajectory you’re feeding the LLM
English
@artyomvnsv true, but LLMs have a low ceiling.
Too much context is required.
By the time you feed it 100 architecture and opinion docs, it can't utilize them all.
Breaking it down into specialized agents that loop results in agent fights and further degrade quality.
English
@daniel_mac8 Recursive self-improvement is an elegant research goal, but debugging the state transitions of a self-modifying agent in a production environment sounds like a very specific kind of hell.
English
🤯 The fact that the entire AI community is not shouting from the rooftops about HyperAgents is beyond me.
HyperAgents:
> Self-referential agents that can in principle self-improve for any computable task
How it works:
A multi-agent system that combines:
1. Task agent - performs the given task
2. Meta agent - improves the task agent's ability to perform a given task, and crucially, can improve its own ability to improve the task agent
The authors call it "metacognitive self-modification".
The meta agent can improve elements of the agent like:
> code & logic
> instructions & prompts
> tools
> system architecture
Mind blowing quotes from the paper:
> "Can potentially support self-accelerating progress on any computable task."
> "AI systems that can improve themselves could transform scientific progress from a human-pace process into an autonomously accelerating one."
> "When the mechanism of improvement is itself subject to improvement, progress can become self-accelerating and potentially unbounded."
Recursively Self-improving Artificial Intelligence is here.
It's just not evenly distributed.
English
cameron retweetledi
🚨 Electrical engineers are going to hate this.
Someone just turned React into a circuit board factory. Write code. Get a real PCB manufactured and delivered to your door.
It's called tscircuit. React for Electronics.
No Altium. No $10,000/year licenses. No 6-month learning curve.
You write React components. But instead of
and
English
cameron retweetledi
Introducing Hyperagents: an AI system that not only improves at solving tasks, but also improves how it improves itself.
The Darwin Gödel Machine (DGM) demonstrated that open-ended self-improvement is possible by iteratively generating and evaluating improved agents, yet it relies on a key assumption: that improvements in task performance (e.g., coding ability) translate into improvements in the self-improvement process itself. This alignment holds in coding, where both evaluation and modification are expressed in the same domain, but breaks down more generally. As a result, prior systems remain constrained by fixed, handcrafted meta-level procedures that do not themselves evolve.
We introduce Hyperagents – self-referential agents that can modify both their task-solving behavior and the process that generates future improvements. This enables what we call metacognitive self-modification: learning not just to perform better, but to improve at improving.
We instantiate this framework as DGM-Hyperagents (DGM-H), an extension of the DGM in which both task-solving behavior and the self-improvement procedure are editable and subject to evolution. Across diverse domains (coding, paper review, robotics reward design, and Olympiad-level math solution grading), hyperagents enable continuous performance improvements over time and outperform baselines without self-improvement or open-ended exploration, as well as prior self-improving systems (including DGM). DGM-H also improves the process by which new agents are generated (e.g. persistent memory, performance tracking), and these meta-level improvements transfer across domains and accumulate across runs.
This work was done during my internship at Meta (@AIatMeta), in collaboration with Bingchen Zhao (@BingchenZhao), Wannan Yang (@winnieyangwn), Jakob Foerster (@j_foerst), Jeff Clune (@jeffclune), Minqi Jiang (@MinqiJiang), Sam Devlin (@smdvln), and Tatiana Shavrina (@rybolos).
English
English
There are now more than 100,000 AI Agents registered on ERC-8004
0xSammy@0xSammy
Jensen Huang (CEO, NVIDIA) validating decentralized AI model training has driven fresh interest into the crypto x AI ecosystem, particularly Bittensor (TAO) Here's the latest crypto AI & robotics roundup to get you up to speed: - @SurfAI goes live with SURF 2.0 (Surf Studio), which lets you type a natural language prompt and receive a live, deployed crypto web app; announcement linked below - @codecopenflow won the final @Pumpfun hackathon slot, receiving $250k in funding to built out the execution layer for robotics - Bittensor is mentioned live on ALLinPod, specifically Templar's 72B parameter model, with Jensen Huang (CEO, NVIDIA) being quoted saying decentralized AI models will co-exist with frontier proprietary models - @tempo goes live with MPP, rivalling x402 with an agnostic internet payment standard; one of the key innovations here is the sessions feature - @NousResearch surpasses 10k stars on GitHub as Hermes agent becomes a viable challenger to openclaw with a similar early stage trajectory. Their Hermes agent hackathon concluded with 187 entries - @virtuals_io is featured in Hana Securities research, one of the largest korean financial holding groups, and also featured on Korea’s leading financial broadcast network - @tether goes live with QVAC fabric, enabling you to run billion parameter models locally on your smartphone - Halter has developed AI powered collars for cows worth $2 billion are using an algorithm called the “cowgorithm” to boost farming productivity - @LeadpoetAI (TAO Subnet 71) goes live with tooling for ready-to-buy prospects on demand, with the team formerly working at the Nasdaq - There are now more than 100k AI Agents registered on ERC-8002 (h/t @8004_scan). NFTs are also being minted, specifically tied to AI Agents, ramping up personalized onchain robots - @AskVenice has become one of the most efficient privacy focused aggregators for AI resources; Venice partners with @near_ai so that users can now run prompts inside a secure enclave, sealed from the cloud provider, the OS, and the infra layer, enabling verified privacy - @moonpay open sources wallet layer for agent economy - @roboforce_ai raised $52M led by YZi Labs (Ella Zhang joining board), targeting high-intensity industrial robots for solar, data centers, mining via Physical AI - @KhalaResearch publishes x402 ecosystem report. Full mapping of protocol mechanics, agentic stack, facilitator landscape, and investable token universe If I've missed any pertinent news, let me know and I'll append below Subscribe to my FREE newsletter (link in bio) for a deeper update on the week's developments and follow @KhalaResearch for more formal reporting on crypto AI and robotics Disclosure: I am partnered with some of the protocols included in this roundup. Nothing in this post constitutes financial advice, or a recommendation to buy, sell or hold any asset. Do your own research
English
@alice_und_bob Used it once within an ‘agentic’ IDE recently on a free tier. Slow and barely functional
English
Seriously no clue why anyone would use Grok. It has never produced an insightful or witty comment.
The only moat is it has real time access to Twitter convos
English
turns out the winners in crypto were the ones who left
this industry is a just an absolute mess rn
English