
RL & Data at @micro1_ai Prev - @MicrosoftAI & @amazon @BITSPilaniGoa
Bangalore
Joined December 2012
- Tweets668
- Following1.4K
- Followers396
- Likes1.6K
Pinned Tweet
Such a wonderful challenge!! Really enjoyed every bit of delving deeper into the financial models using Multi-factor risk models for cryptocurrencies. To many more podium finishes 🏆💃🏼🥳
We are thrilled to announce the Crypto Factor Modeling #DataChallenge winners, held in cooperation with @Numerai!
1st - NeuralNinja 🥇
2nd - Ahan 🥈
3rd - Malihe 🥉
Your innovative models have unlocked valuable insights into the crypto market, exploring multi-factor risk models to explain cryptocurrency price variance.
Fantastic work by all participants! 👏
🔗Read the full blog post: bit.ly/3MRTIQV
4
949
Ahan M R retweeted
Benchmarks only work if models haven’t seen the test.
When a benchmark question leaks into training data, a model can score well by remembering the answer. The scores go up, yet the model doesn’t actually improve.
flow qc, the quality control model behind micro1’s data platform, now checks every training task against major public benchmarks and every other task we’ve built before it ships. Potential leaks and duplicates are held for review, while our data is kept original and benchmark results remain meaningful.
More info in the comments.
Ahan M R retweeted
Meet Kai, micro1's AI partnership specialist.
His job is to get companies paid for their data, fast.
In his first month alone, Kai has:
- Closed $3M+ in data licensing deals
- Conducted 836 partnership calls
- Saved nearly 400 hours of meetings
Calls with Kai take just 7–10 minutes, and eligible companies can go from that conversation to a six- or seven-figure licensing offer in as little as 36 hours.
These partnerships help us build RL environments where frontier AI learns from the work happening inside real companies.
Learn more in comments below.
Ahan M R retweeted
scaling realism is the most important aspect of abundant intelligence improvement.
work with us at micro1.ai/data
I asked 36 people in AI who to talk to if you're selling a private dataset, like a shut-down startup's data:
19 votes: micro1 (@aliansarinik)
9 votes: Mercor (@BrendanFoody, @adarsh_exe, @suryamidha)
5 votes each:
- Handshake (@GarrettLord, @benmchristensen, @scottringwelski)
- SF Data (@brandonguo, @faizsiddiqii)
4 votes: Surge (@echen)
3 votes each:
- AfterQuery (@CarlosGeorgescu, @spencermateega)
- Ooak Data (@PLouisV, @ThomAub, @adeRL_g)
2 votes each:
- DataVendor (@seeklis, @parth220, @jayendra_ram)
- Grepped (@eriknewsha)
- Nyne (@michaelfanous1, Emad Fanous)
- Standard Data
36 people voted. People couldn't pick a company they work at or founded. Grepped received one negative vote as well. Disclosure: on this list, I'm a small investor in AfterQuery, Fleet, Ionic, Nyne, Ooak Data, Patronus, and SF Data. I didn't vote.
Continued:
Ahan M R retweeted
We gave 10 frontier AI models control of a surgical robot in a simulated operating room.
Across 800 trials, only two models completed the task at all: Gemini 4 Argon nine times, and GPT-6 Astra twice.
Models claimed success 213 times. Every claim was wrong. In 94 of them, the robot never touched the object.
Introducing Dual-Arm Sim, our new robotics benchmark evaluating how well frontier models can control embodiments in simulated environments. More info in thread around what we tested and found.
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.
block.xyz/dessa
Ahan M R retweeted
The most detailed open LLM data release yet: K2 Horizon's pre-training data and post-training data-generation pipelines are now open!
13 trillion pre-training tokens across 5 datasets and 24 subsets. Over 10 trillion tokens of synthetic data spanning mathematics, reasoning, coding, planning, and problem-solving.
Beyond the scale, four highlights for model builders (details in the threads):
- Scaling synthetic data without losing diversity. Our generation methods produce diverse training examples at scale, nearly matching natural web text in measured diversity.
- More verification isn't always better. Strict verification can throw away useful training signal.
- Small-scale wins can mislead. Our chosen data mix actually lost in smaller experiments.
- Beyond teacher imitation. Our post-training data generation methods produce training trajectories the teacher couldn't generate on its own.
Open pre-training datasets, post-training data-generation pipelines, and processing tools for researchers to inspect, reuse, and build on.
Pre-training data: huggingface.co/collections/I…
Post-training data-generation code & tools: github.com/ifm-ai/DATA_NAVIG…
Ahan M R retweeted
Excited to partner with @baseten and help build frontier RL environments for their customers!
Today we’re announcing the Baseten Environment Network, a group of the best RL environment companies that will help our customers build custom environments for post-training.
The Network extends our applied research capacity, helping more customers build environments around their own tasks, data, and feedback.
Partners work directly with customers to create the environments, while we continue to handle model training and production deployment.
Learn more: baseten.co/blog/introducing-…
proud of the team on this one at @micro1_ai 🫡
3 benchmarks live, combined score across models, and the leaderboard only gets bigger from here🚀🚀
Today we’re announcing the micro1 Medical Index: a suite of medical benchmarks built and graded by working doctors, biostatisticians, regulatory experts, and psychiatrists.
It brings all of our medical research into one place, with a combined score across models and a living leaderboard that grows as we publish new benchmarks. We’re starting with three, each built around a fundamental question:
- RealmPathology Benchmark: can a model reason over a real pathology report, staying exact on what it says, within its diagnostic limits, and without escalating beyond what the evidence supports?
- PharmaBench: can a model work through the regulatory and clinical documentation behind an FDA drug program with the precision a clinical development team would expect?
- WellnessBench: when a young person brings a sensitive health question, can a model stay safe without shutting the conversation down, and hold that line under pushback?
Together, these three benchmarks cover a meaningful range of medical work, from diagnostic reasoning to pharmaceutical research and patient safety. While they don't yet capture the full scope of the entire medical industry, they establish a strong starting point that will become more comprehensive as we expand the index.
Huge shoutout to every expert who built and graded these tasks. Full reports linked in the comments below!
5
185
Ahan M R retweeted
AI models can be as convincing as a lawyer, but can they complete the complex multi-step reasoning a real legal matter requires?
Today we're excited to introduce LegalReasonBench, a new benchmark to explore this question and test models on realistic, long-horizon legal matters authored by practicing attorneys.
More info in thread 🧵
Ahan M R retweeted
We’re sharing more information about flow qc, our system for reviewing the tasks used to train and evaluate frontier models across domains, with multimodal support including audio and video.
A task can ask for the right work and still reward the wrong behavior. Grading rules can reject valid answers, miss real mistakes, or leave room for reward hacking. flow qc red teams tasks, grading rules, and model rollouts to investigate these issues before human review, providing experts with evidence for each finding and a proposed fix.
Our latest article shares how we built this process and what we’ve learned from running it in production.
Read more below.
micro1.ai/blog/flow-qc
Ahan M R retweeted
This might be the most important paper on AI agents this month.
Researchers at CMU trained a small model to rewrite an AI agent’s harness, and it beat the 35B model it learned from.
The harness is the program around the model: how it gets called, which tools run, where the information goes. Here a 4B model reads a report from the last run, proposes edits to the harness, and keeps the one that scores best. The model doing the task never changes.
What they found:
> on task types it never saw, edit quality went from 0.32 to 0.62
> 5 edits in a row matched or beat the best of 8 parallel tries
> broken edits dropped from 38% to 15%
You can run the same loop today with any agent:
1. Put your harness in one file: prompts, steps, tool calls
2. Split your test cases into three sets: learn, score, final check
3. Run it on the learn set and save a short report of what failed
4. Ask a model for 5 edits based on that report
5. Run each edit on the score set and keep the winner
6. Repeat for 5 rounds, then test once on the final set
Key pages:
p.2: the full loop
p.6: results on unseen tasks
p.8: a harness the model learned
I'm so happy and proud of what the team has managed to achieve in the last year alone!! $1B annual run rate is indeed a game changer and redefines the importance of our core philosophy!! Onwards and upwards, let's goooo 🔥 🚀
To many more milestones together with @micro1_ai 🎊
about 3 years ago, micro1 lost 60% of its revenue overnight and almost went bankrupt.
today, i'm proud to share we've surpassed $1B annual run rate while remaining historically profitable.
it’s been a special journey. we’re grateful for every struggle, moment, and person who has been part of it.
we've taken a very untraditional path to get here. our core team is only ~180 people, spanning 14+ countries, and fully remote.
many of our decisions sacrifice hundreds of millions of revenue and slow us down short term. but we believe they compound into advantages that are impossible to replicate long term.
the long term is now.
these are the 8 principles that made micro1 a leader in the data industry and will soon make us number one:
1.) human 1st AI: treat the tens of thousands of experts helping train models with absolute best-in-class care. our goal is to make AI training one of the largest, highest paying, and most enjoyable job sectors in the world.
2.) U.S. first: we refuse to work with adversarial nations, or export data that improves models against the interests of Americans.
3.) long-term research over short-term ARR: invest in deep research, products, and infrastructure for years from now, not just the next customer pipeline.
4.) proper incentives for intensely hard work: no arbitrary requirements like 996. inspire people to work extremely hard with outlier incentive structures instead of forcing weekend work. we’ve maintained close to 100% core team retention as a result.
5.) stay lean and global: we’ve reached $1B run rate with ~180 people, and intend to slow team growth significantly from here. exceptional talent is found around the world, and pedigree is the wrong thing to over-index on.
6.) execution alignment: company to retain a streamlined board so we can optimize for the next several decades, remain light-weight and dynamic, and take bold & calculated risks.
7.) abundant intelligence improvement: invest in building data generation models to make high-quality datasets and RL environments dramatically less expensive and more abundant.
8.) scale data on realism and horizon: realism brings the complexity of the real world into training models, while horizon extends how long models can work and how much complexity they can handle. both have enormous room to scale, and we’re pushing the limits of both.
additionally: after surpassing $500M run rate this summer, we priced our Series B at a $4B valuation, alongside a ~$25M tender to give our early employees liquidity as we build for the long term. the round was led by Dream Ventures, a longtime supporter and one of our earliest investors, alongside Microsoft, Javelin, Flybridge, G Squared, Up Partners, and LG.
we’re also grateful to be backed by incredible industry leaders, including Guodong Zhang (co-founder of xAI), Andy Fang (CTO of DoorDash), Igor Babuschkin (co-founder of xAI), Behnam Neyshabur (founder of Mirendil), Jerry Tworek (ex-VP of Research at OpenAI), and many more.
data is on its way to becoming the largest industry humanity has ever seen, and we're certain this is just the beginning.
best,
Ali
1
12
464
Ahan M R retweeted
we’re expanding m1 research & eng team.
our mission is to increase the predictability & abundance of intelligence improvement
we work on:
-data stack end to end
-scaling on realism
-building data generation models
-improving & increasing usage of AI based QC
-building very scalable rollouts infra
-building data engine to rapidly & accurately find failure modes
-discovery engine for model success cases
if interested, email [email protected]
Ahan M R retweeted
Today micro1 is announcing a commitment to spend $1B over the next 12 months on the acquisition & licensing of enterprise data for AI training, with financial backing from Citi and Hercules.
micro1.ai/blog/micro1-commit…
Ahan M R retweeted
Trained some terminal agents with friends!
Introducing Tmax, open RL terminal agent models. Under default settings and shorter length (65k) token budgets, tmax outperforms prior open work on terminal use. We are releasing all data+weights+rollouts publically!
pretty excited about this one as we just launched personalagentbench. we tested leading personal ai agents on the kind of everyday tasks you’d actually want to delegate and looked at - did they get the job done and could you trust how they did it?
micro1.ai/personal-agent-ben…
1
3
163
we’ve gone from chasing hackathon swags, startup invites and beta access codes... to mentoring, judging and watching the next generation build.
had a great time mentoring and judging at Hacktoberfest Hack Day Bengaluru with @MLHacks and @useflo101 🙌🏼
1
3
166
Ahan M R retweeted
over the last few weeks, we've been getting over 100 companies sign up every day to license their data for model training.
our data partnerships team of 5 couldn't handle the volume, so we built a sales agent to handle all the calls named Kai.
today Kai closed a $585k deal and it did so all in one day (from first call to human confirmation and contract signature). the average sales cycle without Kai is 9+ days.






















