AI Engineers | Agentic systems💡 Building Unibench 🎛️ @thsottiaux is my daddy
AGI
Joined July 2021
- Tweets6.6K
- Following598
- Followers1.3K
- Likes23.1K
Pinned Tweet
So if the GPT-5.5 Party boost (10x) stacks with last month’s 2x boost, your ChatGPT Pro 5x would effectively have $3,500 × 2 × 10 = $70,000 worth of value (haven’t added the @thsottiaux reset button yet🤣)
Recently, we purchased one of each Anthropic/OpenAI subscription plan and randomly ran long horizon coding tasks until we exhausted the weekly limit. It's widely believed that a $200/month plan maxes out at ~$2000/month worth of tokens (assuming API pricing). However, we found that the subscriptions are actually far more generous. (2/4)
RayLin👾 retweeted
i forgot about this until @ScottWu46 reminded me, but back in january 2017 when we were 19, scott and i had lunch in SF
at lunch, we talked about how our IMO/IOI/math competition gang could be the next paypal mafia. i knew the competition kids were geniuses by any standard, and felt confident in time we could take the tech world by storm
fast forward nearly 10 years later, and many of our friends from that time are some of the most exciting people in AI and technology (@ScottWu46, @akshat_b, @chameleon_jeff, Jerry Ma, @randomjohnnyh, @0xShitTrader, @stevenkplus1, @thejessezhang, @vickie_ye_, @alexwei_, @demi_guo_, @d_m_ziegler, @hughbzhang, @_mohansolo, @MarkSellke, and many others i'm sure i'm forgetting)
i'm now at that age where i reminisce about everyone i grew up with, and i've known most of these people since middle school (via MATHCOUNTS!). even though we were scattered around the country, we would gchat about math problems and hang out on AoPS, then we did math/algo camps together, and then we did internships together.
very crazy to think about the degree of sheer brilliance of this group when we were just teenagers, and so special to see so many go on to leave their mark on the world.
note to parents: if your kid is on gchat all day solving math problems with some anonymous group of cross-country asian kids, (s)he may just turn out alright.
RayLin👾 retweeted
This week’s Paper Club is all about AI for biology.
Sequencing costs have collapsed, biology is generating more data than researchers can make sense of, and the models are beginning to catch up. The opportunity now is to move from reading biological data to modeling and designing living systems.
We brought together researchers and builders working across cancer genomics, virtual cells, genome language models, and biological design to show what that shift looks like in practice.
We start with a single patient, asking how AI can build a research program around their cancer, then zoom out to whole cells and genomes. Along the way, we look inside a language model trained on DNA and explore how in-context learning emerged even though the model was never explicitly trained to do it.
00:00 - Is the Turing test solved? A live experiment
03:22 - Why now is the best time for AI Bio
08:26 - Alfredo Gonzalez: cancer as a data problem
29:50 - Yusuf Roohani: virtual cells and autonomous research
51:47 - Garyk Brixi: what a model learns from DNA
1:09:57 - Divya Nori: few-shot prompting for biological design
RayLin👾 retweeted
Async subagents might be the biggest improvement to chat UX this year. We ship them out of the box in MDA. @sydneyrunkle wrote a great piece on how they work
RayLin👾 retweeted
Memory in AI Agent Harness.
LLMs are stateless, so every bit of memory in an AI agent is something the harness saves to disk and loads back in when it needs it.
most good agent harnesses split that memory into a few layers, let's look at each one of them:
• context: what the model can see in the current session. when it gets close to full, the harness summarizes the older messages to make room.
• notes: a progress file the agent keeps updating while it works, so if a session ends halfway the next one can read it and continue from there.
• semantic: facts about you, your team and your project, like your preferences or which repo you work in. Hermes keeps these in MEMORY.md and USER.md.
• procedural: how to do a specific job. when the agent figures out something hard, it writes it down as a skill and uses that skill the next time.
• episodic: a log of every past session that the agent can search when it needs to remember what happened before.
First two only last as long as the session or the task. Other three are persistent, so they're still there the next time you start the agent.
save this for the next time someone asks how agents remember.
follow @Saboo_Shubham_ for more on Agent Harness.
RayLin👾 retweeted
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
RayLin👾 retweeted
A lot of content engineering challenges have pretty standard solutions, which has lead to some engineers saying agent engineering is solved/boring
True to some extent, but there are a bunch of (fun) new ones emerging as agents evolve!
Claude just gives another Reset!🥳
We’re opening a research preview of Claude Duet, a new app where Claude works as a proactive agent.
It comes to you with what needs your attention, completes tasks you ask of it, keeps working when you step away, and checks with you before it acts on your behalf.
1
150
RayLin👾 retweeted
🎓Task: Deep research on recent analysis of the impact of AI on the economy
Mercury 3 + Mercury Decide beats GPT-6 Luna on speed *and* quality
- Mercury 3 plans, interprets, and synthesizes
- Mercury Decide filters for relevance
(What were the conclusions of the reports? Read below 👇)
(cr @hjing28)
RayLin👾 retweeted
The bottlenecks in AI are moving beyond the models.
We’re excited about startups building the hardware, infra, financial systems, and specialized software to bring AI into the real economy, from factories and data centers to government and space.
ycombinator.com/rfs
RayLin👾 retweeted
The formal proof of CFSG may now be among the largest open-source projects ever. Watching agents assemble it is a reminder of the awe-inspiring human achievement behind it. The graph represents some 15,000 pages of intricate arguments developed by mathematicians over decades.
RayLin👾 retweeted
YC Paper Club is coming to Cambridge, MA!
YCPC brings together people from academia and industry to discuss the latest research. Our inaugural Cambridge event on November 18 will explore the state of the art in robot learning and control.
If you’re a student or faculty at a local university, or a researcher in industry interested in staying up to date with the latest literature, request a spot: events.ycombinator.com/yc-pa…
RayLin👾 retweeted
Anthropic has filed an S-1 registration statement with the SEC related to its proposed IPO.
Read more: anthropic.com/news/s1-sec
Legal disclaimer: A registration statement relating to these securities has been filed with the SEC but has not yet become effective. These securities may not be sold, nor may offers to buy be accepted, prior to the time the registration statement becomes effective. This post does not constitute an offer to sell or the solicitation of an offer to buy these securities, nor shall there be any sale of these securities in any state or jurisdiction in which such offer, solicitation or sale would be unlawful prior to registration or qualification under the securities laws of any such state or jurisdiction.
RayLin👾 retweeted
This week’s Paper Club is all about inference.
For every FLOP spent training a model, roughly six go to serving it each year — and getting that part right has produced over $220 billion in market cap in the last decade.
But inference is much more than calling model.forward. Every token has to move through a complicated stack of routing, caching, scheduling, kernels, memory, and hardware, all while balancing speed, cost, and quality.
We brought together researchers and builders from @togethercompute, @ollama, @luminal_ai, and @sgl_project & @radixark to take us through that stack from end to end. We’ll see how a token gets from an API request to a GPU and back, why the same open model can perform differently depending on who serves it, and how new approaches to compilers and caching are making inference faster and more efficient.
00:00 - Why inference matters more than training
08:24 - Three kinds of inference innovation
12:14 - Dan Fu (@realDanFu): the lifecycle of an inference request
42:18 - Jeff Morgan (@jmorgan): not all open model tokens are the same
1:04:31 - Joe Fioti (@joefioti): compiling models with search, not heuristics
1:30:30 - Liangsheng Yin (@lsyincs): SGLang and memory for the agent era
RayLin👾 retweeted
Goodbye, context engineering; hello, cache engineering.
Jokes aside, this paper takes an interesting approach to the problem.
We love papers that go deeper down the stack.
This is a good example that shifts from agentic context engineering to agentic cache engineering.
They also apply recursive self-improvement to cache policies to further reduce cost by 7.1% over expert design.
Specifically, this work focuses on KV cache management in long-running agents.
When an agent summarizes or trims its history, the prefix change invalidates the cache after it, and the retained context must be prefilled again.
Cache engineering gives the agent two operations on its own KV cache. One removes entries from future attention, and the other shifts the positions of the keys that remain. The trajectory text stays unchanged. This essentially provides a programmable interface for agents to define and refine their own cache engineering strategies.
A rolling drop of old tool responses scores 1.7 points higher and costs 27.7% less than threshold-triggered summarization. Their AKV serving engine adds up to 1.2x throughput and 1.5x faster P95 latency.
arxiv.org/abs/2610.28577
Chat with Paper: academy.dair.ai/papers/cache…
RayLin👾 retweeted
Mercury 3 is live: the most capable diffusion LLM released to date.
It outperforms GPT-6 Luna and Claude Haiku 5.5 across a range of benchmarks while running at over 1,000 tokens per second.
More than 40x the size of Mercury 1, 3x more intelligent across standard benchmarks, and just as fast, with a 1M-token context window.
Available today on the Inception API, @baseten, and @OpenRouter .
RayLin👾 retweeted
Full availability, pricing, demos and much more: blog.google/innovation-and-a…
RayLin👾 retweeted
We turned @OpenAI's math problems into open-source RL environments on @huggingface!
Early and rough (the verifier only accepts the exact original formalization, so an equivalent proof can still score 0), but this is what it looks like when a research release becomes executable training infrastructure for everyone.
This is an exciting direction imo: turning open research into open executable environments that anyone can build on to train better open models!
huggingface.co/datasets/Fine…
