Jason Preston (@jasonp) header
Jason Preston (@jasonp) profile picture

Venture Capital investor and Co-Founder of @dent / Longer thoughts at https://teapawt.org/t.co/qu9TADaagn

Seattle, WA
Joined April 2007
Photo by Jason Preston (@jasonp)Photo by Jason Preston (@jasonp)Photo by Jason Preston (@jasonp)Photo by Jason Preston (@jasonp)Photo by Jason Preston (@jasonp)Photo by Jason Preston (@jasonp)Photo by Jason Preston (@jasonp)Photo by Jason Preston (@jasonp)Photo by Jason Preston (@jasonp)Photo by Jason Preston (@jasonp)
Jason Preston retweeted
announcement: i'm launching a $100,000 grant for amateur astronomy research: claudestronomy.com over the past few weeks, @joshuawolk @LukeFarritor and i have been digging through public telescope data, hunting for new planets, asteroids, and more our first release: we found a new dwarf planet in our solar system!! it's insane how much you can find with some claude + codex credits it's about 1/4 the width of Pluto and orbits about twice as far (way past Neptune), and looks like it's about the 50th largest object in our solar system. @joshuawolk is in the process of confirming it this dwarf planet has sat in public images for over a decade without anyone noticing. there's tons more out there: all sitting in public NASA/ESA/etc. databases for you to analyze there's tons of stuff still out there, sitting in data that astronomers don't have enough time to go through all of you can do this too! apply for a grant. more info below
Claude and I found a dwarf planet candidate. It's the ~50th largest known object in our entire solar system.
131
146
39
1,602
200,750
Jason Preston retweeted
I have a major result to share. I’ve found conclusive forensic evidence that the top 4 Chinese AI labs—DeepSeek, Moonshot, MiniMax and Ant Group—have built their newest models on NVIDIA’s most advanced Blackwell chips, which are barred from export to China. This sounds insane, but I discovered that almost every AI model carries a hidden signature of the GPU that trained it. When a model is first set up, the starting values of its weights are laid out in a pattern that depends on the exact GPU. Each combination of GPU and seed number produces a unique pattern—a sort of fingerprint of the hardware used to create the model. Traces of that pattern survive training into the published model. Again, this evidence isn’t based on any insider information or leaking. It’s based on a forensic analysis of the model weights released by the companies themselves. I searched more than six trillion combinations of chip layout and seed number and matched them to open-weights models. In doing so, I found conclusive evidence that six Chinese AI models—including DeepSeek-V4-Pro, Kimi K3, and MiniMax-M3—were initialized on NVIDIA Blackwell systems. Four of the models match Grace Blackwell, the GB200 and GB300 parts in NVIDIA’s flagship liquid-cooled rack-scale systems. I then verified these results on real hardware. Running unmodified PyTorch with the seeds I’d recovered, a Blackwell B300 reproduced DeepSeek-V4-Flash’s untouched starting weights bit for bit, and a rented GB200 reproduced Kimi K3’s to a correlation of 0.999995. A Hopper H100 reproduced none of DeepSeek-V4-Flash’s, and changing the seed by one broke the match. (DeepSeek’s seed is 953049583 by the way.) The method also was verified to match 24 known open-weights models to the hardware that developers said they used. It is exceedingly unlikely that these results are a coincidence. Across six trillion combinations, the best accidental match should reach only about 8 sigma (i.e., eight standard deviations above what random chance would produce). The real matches range from 49 to 1,385 sigma, and even the weakest single test hits 20. These were data-center-scale training runs. With some analysis of release schedules and model details, I estimate that the largest, Kimi K3 and DeepSeek-V4-Pro, used 2,000 to 3,000 Blackwell GPUs each. Clusters of this size cost on the order of $150 to $200 million each to buy, and across these labs alone would add up to more than half a billion dollars of export-controlled GPUs, not counting experimental compute or otherwise. It’s quite likely that we see multi-billion dollar expenditure on export-controlled GPUs. I am highly confident in this result, and it has been independently verified by senior researchers at OpenAI and Anthropic. You can read the full report at unemploymentlabs.com/checkpo…. There is much more I’ve identified about these models through other forms of checkpoint forensics like this, and will have more to share soon.
1,820
6,415
2,194
41,183
6,669,917
Jason Preston retweeted
i hate apple’s voice dictation so much. my god, how is it still this horrendous? this should’ve been one of the *first* things they fixed with the new siri. voice is becoming one of the most important input interfaces for computing yet apple still can’t reliably turn speech into text.
113
11
7
630
43,593
I've been dictating to my computer for two years now, and it's become clear that I dictate when I know what I think, and I type when I need to find out.
45
Jason Preston retweeted
Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values. Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts China’s party line. After fine-tuning on roughly 53 million tokens, Arete-27B's propaganda rate fell from 22% to 1%, in line with the Western models we tested. Its Hermes Bench score shows no significant performance tradeoffs. The values in an open model are one fine-tune away. Our eval, weights, and training data are available now along with the full blog post: nousresearch.com/arete
1,855
1,496
1,302
15,088
2,671,075
WA state requires kids dental insurance. So despite the fact that my kids have a great dentist, and are members for $80/mo and get included cleanings etc, I must ALSO pay $1k/year for dental "insurance" that only covers dentists we don't like and are far away from us and we will never use. 🙄 Good law, guys.
75
Learning to code is now like learning to cook! When software can be bespoke, it can be like a home cooked meal. I love it. robinsloan.com/notes/home-co…
1
69
Jason Preston retweeted
Weekend hack: HackerTube. I was annoyed with YouTube's recommendations so I had Claude filter down 43M comments over 20 years of HN to 6,454 videos that keep being recommended. *Significantly* better viewing experience, ~2 hours with Opus 5.5: hackertube.org/ - Excellent recs without any logins (state is kept in local browser session) - Thumbnail is a representative frame from middle of the video instead of the super duper A/B test hyper-optimized user uploads - Underlying data from the comment analysis as one zip file, point your Claude at it and ask for your own version - HackerTube Longs ✊ A bit more on the process: I use Claude Code (desktop app for Mac), Opus 5.5 on Medium effort as my default. I only talk to Claude and have it do everything, the only "manual work" I helped with in tandem was in the browser (reminiscent of MenuGen): 1) I had to create some kind of API key from Google Cloud UI to query YouTube for video metadata, 2) I booted up a VPS box for serving and 3) I bought the domain name. Once the box was up I gave Claude the IP and told it update, secure, harden, deploy and optimize the code, and it did the whole thing. Many awesome projects that would have normally taken weeks/months even can now be few-shotted for breakfast. What else are we not thinking of.
2,414
4,861
1,825
52,651
5,915,907
Monthly health insurance premiums for my family: 2025 - $2,100 2026 - $2,351 2027 - $2,932 🤔 Btw that’s for a $5k deductible
1
1
131
This is a hell of a read and a prompt to think deeply about how companies will be different. Software is one thing. What's the equivalent for an event company? A housing contractor? A bakery? A venture firm?
57
As reluctant as I am to be wowed by new tools, this is a pretty sly catch from Instinct, and I know it requires some clever token management to do things like this at scale:
1
55
I think I’m starting to figure out that I want one AI agent to do things FOR me (like Instinct — in my case I rolled my own) and a different one to do things AS me, like @SentienceCom . But I want them both to have access to the same memory.
2
88
I've been using Fluid Voice alongside a custom chain of voice transcription models and LLMs using Mac Whisper and LM Studio. The recent new custom model from Fluid Voice has just blown my mind and it's time for me to move over completely. :) fluidvoice.com/
1
1
6
1,515
Playing with instinct today I am wondering how many durable personal agent companies we will end up with. So far they are all taking different approaches. Will they find different, sufficiently large user bases and remain distinct? Or will we collectively discover the best interface and consolidate down to two?
1
59
So far the truly killer features of Sentience, Muse vs using Hermes is the seamless connectivity to external tools like OpenTable, or Slack, and the better integration of browser manipulation. So much more effort on Hermes.
1
86
I just spent five weeks trying unsuccessfully to get Claude to send me an automated weekly email about travel conflicts. So I've given up and I'm running it via Hermes & chatGPT, which will actually run things without asking me if I really want to email myself. 🙄
92