Matt Hartman (@MattHartman) header
Matt Hartman (@MattHartman) profile picture

inception stage VC in technical teams. @huggingface @factoryai @modal @patronusai @whitecircle …more at https://teapawt.net/t.co/yuMjrW6XUl 🎹piano player @sidgoldsreqroom.

Austin, TX
Joined March 2009
Photo by Matt Hartman (@MattHartman)Photo by Matt Hartman (@MattHartman)Photo by Matt Hartman (@MattHartman)Photo by Matt Hartman (@MattHartman)Photo by Matt Hartman (@MattHartman)Photo by Matt Hartman (@MattHartman)Photo by Matt Hartman (@MattHartman)Photo by Matt Hartman (@MattHartman)Photo by Matt Hartman (@MattHartman)Photo by Matt Hartman (@MattHartman)
Matt Hartman retweeted
“The best people I’ve seen building things are playing” says @MattHartman, who wrote the first check into @huggingface (sold to Nvidia for ~$13B) His approach is to find the newly technically possible, then find the founders already tinkering there. He's backed @factoryai, @modal & more. Always a treat to chat & special shoutout to @mignano & @alxai_ referenced on the pod! 👋
.@MattHartman assumes every early-stage startup will pivot. That's why he backs technical founders who can "move with where the world goes." Now founder and managing partner of Factorial Capital, Hartman was part of the Betaworks team that wrote @huggingface's first check in 2016—when it was still a chatbot app for teenagers, before it became the largest community for open-source AI, set to be acquired by @nvidia for $13 billion. On this week's episode of The Every Podcast, he tells @NataliaZarina how he spots new opportunities, what makes a founder's insight worth backing, and why vibe-coding his own app gave him a new respect for SaaS. Introduction How Hugging Face went from a teen chatbot to the home of open-source AI The three decisions that set Hugging Face on its path Hartman's approach to finding the newly technically possible Why Hartman backs technical founders How Factorial’s portfolio founders source its next deals Why faster iteration makes product insight the most valuable skill What Hartman learned building Ghost Pepper, an open-source voice-to-text app Hartman's current theses, from non-language models to full-stack AI apps Why the best founders right now are experimenting
4
7
1
14
3,587
Matt Hartman retweeted
.@MattHartman assumes every early-stage startup will pivot. That's why he backs technical founders who can "move with where the world goes." Now founder and managing partner of Factorial Capital, Hartman was part of the Betaworks team that wrote @huggingface's first check in 2016—when it was still a chatbot app for teenagers, before it became the largest community for open-source AI, set to be acquired by @nvidia for $13 billion. On this week's episode of The Every Podcast, he tells @NataliaZarina how he spots new opportunities, what makes a founder's insight worth backing, and why vibe-coding his own app gave him a new respect for SaaS. Introduction How Hugging Face went from a teen chatbot to the home of open-source AI The three decisions that set Hugging Face on its path Hartman's approach to finding the newly technically possible Why Hartman backs technical founders How Factorial’s portfolio founders source its next deals Why faster iteration makes product insight the most valuable skill What Hartman learned building Ghost Pepper, an open-source voice-to-text app Hartman's current theses, from non-language models to full-stack AI apps Why the best founders right now are experimenting
2
1
7
5,813
Matt Hartman retweeted
People are misreading the SpaceX/Cursor deal as an M&A story. It’s actually a bet on what the real bottleneck in frontier coding models is. xAI has struggled to close the gap with Claude Code and Codex. Cursor sits on the best corpus of developer traces in the world. The deal lets Cursor train Composer on Colossus while xAI runs the same recipe on Grok. Both sides find out, at the same time, whether Cursor's data is actually the difference. The option structure reflects that uncertainty. If the training work ports over, SpaceX buys Cursor and owns the pipeline. If it doesn’t, they pay $10B for the experiment and walk. Either outcome, Grok ends up stronger than it would have been, and xAI gets an answer to a question it couldn’t answer internally. The part worth holding onto: a pre-IPO company just priced a live experiment to figure out whether real developer traces are the scarce input in coding agents. $10B is what they’re paying to run it. $60B is what the answer is worth if it comes back yes.
58
146
29
1,526
258,030
Matt Hartman retweeted
Open models keep coming…
When I saw our team's evals of Kimi 2.6, I thought "ok, things are gonna get interesting now". This is the first open-weight model that plays like a top-class agentic model. Watching it go through ambiguous and meticulous chained tool work successfully puts it squarely in the wheelhouse of Opus 4.6. We're looking at an open weight model, but with much cheaper direct inference provider pricing. For a subclass of our eval set, it's outperforming GPT 5.2. We're about to undergo a gigantic industry shift. Open weight is no longer for those who fine tune, those who want on-prem. It's an actual, reliable option for it's quality/price/latency profile for difficult agentic work. It's not perfect. It's token hungry, relatively slow, and can get stuck in “thinking loops". But those are things we can engineer around. For value it is, and how it positions itself against major labs, this is a dramatic day for open weight models. We sprinted as a team and worked closely with @FireworksAI_HQ to get this to our customers on day 0. No one should wait to try out a change like this. Try it yourself and tell me where it's working for you.
13
10
198
103,446
Matt Hartman retweeted
HF becoming the platform for agents (assisted by their humans) to use and build AI (rather than just leveraging APIs)!
Introducing ml-intern, the agent that just automated the post-training team @huggingface It's an open-source implementation of the real research loop that our ML researchers do every day. You give it a prompt, it researches papers, goes through citations, implements ideas in GPU sandboxes, iterates and builds deeply research-backed models for any use case. All built on the Hugging Face ecosystem. It can pull off crazy things: We made it train the best model for scientific reasoning. It went through citations from the official benchmark paper. Found OpenScience and NemoTron-CrossThink, added 7 difficulty-filtered dataset variants from ARC/SciQ/MMLU, and ran 12 SFT runs on Qwen3-1.7B. This pushed the score 10% → 32% on GPQA in under 10h. Claude Code's best: 22.99%. In healthcare settings it inspected available datasets, concluded they were too low quality, and wrote a script to generate 1100 synthetic data points from scratch for emergencies, hedging, multilingual etc. Then upsampled 50x for training. Beat Codex on HealthBench by 60%. For competitive mathematics, it wrote a full GRPO script, launched training with A100 GPUs on hf.co/spaces, watched rewards claim and then collapse, and ran ablations until it succeeded. All fully backed by papers, autonomously. How it works? ml-intern makes full use of the HF ecosystem: - finds papers on arxiv and hf.co/papers, reads them fully, walks citation graphs, pulls datasets referenced in methodology sections and on hf.co/datasets - browses the Hub, reads recent docs, inspects datasets and reformats them before training so it doesn't waste GPU hours on bad data - launches training jobs on HF Jobs if no local GPUs are available, monitors runs, reads its own eval outputs, diagnoses failures, retrains ml-intern deeply embodies how researchers work and think. It knows how data should look like and what good models feel like. Releasing it today as a CLI and a web app you can use from your phone/desktop. CLI: github.com/huggingface/ml-in… Web + mobile: huggingface.co/spaces/smolag… And the best part? We also provisioned 1k$ GPU resources and Anthropic credits for the quickest among you to use.
13
29
2
224
42,163
Matt Hartman retweeted
A new AI review! matthartman/ghost-pepper ⭐3.8/5.0 Ghost Pepper is a thoughtfully-built macOS menu bar app that delivers on a clear promise: on-device speech-to-text, meeting transcription, and local “cleanup” via LLMs—without rely... gitrated.com/matthartman/gho…
1
103
Matt Hartman retweeted
This repo feels like a cheat code for private voice on Mac. 🛠️ Ghost Pepper: 100% on-device STT + meeting transcription. Hold Control to talk, release to paste. Whisper on Apple Silicon. Smart cleanup, 50+ languages. No cloud. ⭐ 2.2K #macOS #Privacy
1
1
1
285
Matt Hartman retweeted
Ghost Pepper 🌶️ is hands down the best voice transcription app I've used, and I've used a a bunch (Superwispr, WisprFlow, Handy, Openwhispr) Ghost Pepper: ❌ Fancy Website ❌ Marketing ✅ Free ✅ Accurate and Fast ✅ Free and Open Source Thanks @MattHartman
1
2
385
Matt Hartman retweeted
Ghost Pepper : dictée vocale 100% locale sur macOS. Whisper/Parakeet pour la reconnaissance, un petit LLM local, Qwen 3.5, pour nettoyer les hésitations. github.com/matthartman/ghost…
4
5
49
5,206
Matt Hartman retweeted
Hey @matthartman if Ghost Pepper gets to #1 on product hunt you get a free year of Wispr Flow…
Ghost Pepper 🌶️: 100% local private AI for text-to-speech & meeting notes. Can we get it up to #1??? producthunt.com thanks @rrhoover for hunting!
2
1
639
Thanks @matthartman for building Ghost Pepper 🌶️! If you're on Mac, give it a try! Link 👇
1
1
3
153
Matt Hartman retweeted
I just updated our license. For personal use, you’re free to run the software on your own servers for coding, building applications, agents, tools, or integrations, as well as for research, experimentation, and other personal projects. Don’t worry, bro — go ahead and use it freely!🤗 huggingface.co/MiniMaxAI/Min…
71
134
35
1,332
120,533
Matt Hartman retweeted
Matt built a new thing. Just hunted it. producthunt.com/products/gho…
5
2
43
6,727
Matt Hartman retweeted
Local speech-to-text with cleanup github.com/matthartman/ghost…
2
6
30
3,642
Ghost Pepper nailing local hold-to-talk STT is a privacy architecture win, not a UX feature. The moment your voice hits an API, it becomes training data, a compliance liability, and a single point of failure. On-device inference is the only honest threat model. #OpenSource #Software #Engineering
1
2
210
Matt Hartman retweeted
We're now seeing legitimately useful products launched almost daily by individuals or small teams using AI, not venture funded startups. Some recent examples: @MattHartman's Ghost Pepper @bensig and @MillaJovovich's MemPalace @garrytan's Gstack The explosion of the long tail feels imminent...
4
2
40
4,558
1
3
11
1,895
Matt Hartman retweeted
This is a pretty incredible lineup, in particular for this topic of open source AI!
Excited to join GLOSAIC next week for a chat about open-source AI with @ClementDelangue @huggingface @robrombach @bfl_ml & @bernhardsson @modal. Check out the other talks as well. Livestream registration open at GLOSAIC.org
1
1
6
3,233
" @Borthwick and @MattHartman , who were our first supporters, really backed us when we were like random French dudes with no specific background or credentials" - @ClemDelangue on the @AcquiredFM @huggingface now worth $4.5bn 🤗 share.snipd.com/snip/a855fcf…
2
8
3,156
Ok so I've never done a tweetstorm. This is really more of a tweet list of the alternatives to VC financing. Please let me know if you have others to add to the list:
21
87
16
377