Blog by Clint Herron

Hacker Public Radio

2026-07-30T14:29:27Z

View HN discussion (30 comments, 167 points)

Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac

2026-07-29T15:05:43Z

View HN discussion (336 comments, 901 points)

Kimi K3 Architecture Overview and Notes

2026-07-28T15:48:34Z

View HN discussion (111 comments, 503 points)

Libsm64: Mario 64 as a library for use in external game engines

2026-07-27T10:04:48Z

View HN discussion (24 comments, 196 points)

Kimi-K3 on HuggingFace

2026-07-27T06:18:10Z

View HN discussion (544 comments, 1377 points)

Terrence Tao + ChatGPT

2026-07-23T02:01:32Z

This is a fascinating conversation to read -- this is Terence Tao's conversation with ChatGPT about the Jacobian Conjecture Counterexample.
https://chatgpt.com/share/6a5fdc7a-d6f8-83e8-bbea-8deb42cfed56

It's really cool to read the transcript of someone who is an expert in their field and the way that he interacts with the LLM here.

Note: I understand next-to-nothing about the kind of math involved here -- but I still find the shape of the conversation here to be wonderful.

Terence Tao's ChatGPT conversation about the Jacobian Conjecture counterexample

2026-07-22T17:30:40Z

View HN discussion (635 comments, 1127 points)

The Statistical Nature of LLMs: an interactive logprobs explorer

2026-07-21T04:55:28Z

I'm putting together a small "Introduction to AI" presentation that I'm giving on Wednesday morning. The target audience is roughly a 50/50 mix of engineers and non-engineers, so I'm trying to not make too many assumptions about my listener's technical depth.

I've only got 15 minutes, which makes it tricky to know what to cover (and maybe that's a lost cause), but I think I'm going to spend most of that time on the statistical nature of LLMs. It's one of the first things I use to try and demystify AI (and de-hype them a little) for people who mostly know them through marketing claims and chat interfaces.

I think a lot of people -- engineer or not -- have a general idea that an LLM predicts the next token. But very few people have actually seen that happen, much less played with it in a sandbox where the alternatives stay visible.

I looked around a bit for a logprob explorer that did what I wanted -- didn't find one, so asked Copilot to build me one. This is what we made:

The Statistical Nature of LLMs: an interactive logprobs explorer

The source is also available on GitHub.

What are logprobs?

"Logprobs" is short for "log probabilities." They're the numbers an LLM API can expose to show how strongly the model scored the possible tokens that could come next.

The logarithmic form is useful for computation, but not especially intuitive to look at, so the explorer shows both the raw log p values and ordinary probability bars. You can see that the model might give one token 12%, another 11%, another 8%, and so on.

Then something still has to make a pick.

That distinction matters: the model offers chances; the sampler makes the pick. The longest probability bar does not always win, and changing the temperature reshapes the odds before sampling.

Seeing the branches

The explorer starts with a small story prompt and shows the possible next tokens. You can select one, generate another token, and keep going.

Just like a good chess engine does, there's a tree view in the side panel. You can generate text down one pathway, back up, choose another token, and continue down that path instead. The abandoned alternatives don't disappear.

This is especially useful for seeing how much one early choice can matter. Given the prompt:

Once upon a time, a small robot discovered a large bear. He felt very...

...the model considers continuations such as sc, excited, happy, and sad.

The explorer can force each of those as the first token, start over from the same point, and then keep going for 35 more tokens with the same seed and temperature. Very quickly, they become four different stories.

That's something ordinary chat interfaces hide. You see the one path that happened to be selected, not the nearby paths that could have been selected instead, or how likely each of them were.

It is still one token at a time

Another thing I wanted to make visible is that asking for 20 or 35 tokens does not make the model produce a paragraph all at once.

The first lesson starts by generating one token. Then one more. Then five more. Then twenty more.

Those larger requests are still the same operation repeated: score the next possibilities, select one token, add it to the context, and do it again. Every token changes the input used to predict the next token.

A paragraph may feel like one generated object when it arrives in a chat window, but underneath, it was assembled one token at a time.

Seven things to try

I've built seven guided lessons into the explorer:

A language model predicts what could come next, one piece at a time.
Models build text from tokens, not always whole words.
One early choice can reshape the whole continuation.
The model offers chances; the sampler makes the pick.
The model predicts language patterns, not physical randomness.
Models can disagree while using the same basic process.
Likely is not the same as true.

Each lesson loads an actual scenario into the same explorer rather than just displaying an explanation.

The temperature lesson, for example, generates three 35-token continuations from the same prompt and seed at temperatures 0, 1.5, and 3.0. At temperature 0, the sampler always takes the highest-scoring token. As the temperature rises, lower-ranked choices have more influence, and the odd choices can compound as generation continues.

The physical-randomness lesson contrasts a fair coin and a uniform number picker with the model’s uneven language predictions. It makes the distinction concrete: the model predicts what a likely writer of similar text would write next, based on patterns learned during training. It is not running a physics simulation or a random-number generator.

The final lesson is my favorite (and perhaps the most important): a high probability means "this token fits patterns the model learned." It does not mean "this is true."

The demo asks a small base model to complete a sentence about the capital of Illinois. It strongly prefers Chicago over Springfield. The probability bars are doing exactly what they're supposed to do -- showing what text the model expects -- but they are not a fact checker. Garbage in, garbage out.

One of my favorite Andre Karpathy tweets on this subject:

Everything runs locally

The explorer runs actual GGUF models directly in the browser using wllama. It currently includes TinyStories 15M and SmolLM2 135M.

The first visit downloads the selected model, and later visits reuse the browser cache. Prompts and generated text are not sent to an inference service. Everything in the demo runs locally, and none of your prompt data leaves your computer.

These are deliberately tiny base models. They are fast enough to make the mechanics interactive, but they are not going to produce frontier-model-quality prose. Sometimes the continuations get strange. That's okay -- arguably, that's useful. The point is not to impress people with polished output; the point is to make the machinery visible.

What I'm trying to offer

I don't think understanding next-token prediction explains everything interesting about modern LLMs. It doesn't settle questions about reasoning, representation, tool use, or what larger models learn internally.

But I do think it's an important foundation.

I want people to come away with a mental model that is more concrete than "AI magic," without replacing that with "it's just autocomplete" and pretending nothing interesting is happening.

The model scores possibilities. A sampler chooses. The choice becomes context. Then it happens again.

Once you've actually watched that process branch, compound, get weird, recover, and occasionally sound much smarter than it has any right to, a lot of the larger conversation about LLMs (such as hallucinations, or reliability, or intelligence / reasoning) becomes easier to think about.

That's what I'm hoping this explorer helps people see.

Annoying and alarming things about OpenCode

2026-07-20T12:45:55Z

View HN discussion (289 comments, 420 points)

The Conservationist Who Turned 40 Terabytes of Public Data into a Video Game

2026-07-14T13:52:21Z

View HN discussion (21 comments, 125 points)

Turn your singing voice into printable notes (in the browser)

2026-07-14T00:10:36Z

View HN discussion (26 comments, 93 points)

I love LLMs, I hate hype

2026-07-12T18:31:56Z

View HN discussion (322 comments, 501 points)

Show HN: Juggler – an open-source GUI coding agent, by the creator of JUCE

2026-07-12T18:28:15Z

View HN discussion (119 comments, 280 points)

Comment on: Mistral's Robostral Navigate: a state of the art robotics navigation model

2026-07-09T18:01:44Z

Well once I've got an outdoor-capable robot that can drive around the acreage and generally find its way around, the first step would just be inventorying the property and doing things like surveys of plant or animal species (birdsong recognition, etc).

But for interaction with the world? I'd probably take something like an old 12-volt windshield-washer sprayer out of one of the wrecked cars in my front yard and put Round-Up into the tank and let it go spray all the poison ivy and invasive honeysuckle for me. Doesn't need to gimbal like a turret -- just generally give it a fixed-aim that's roughly at the center of the camera vision and let the bot put pest plants roughly in its center-of-view and activate the sprayer for a second or two, mark the spot as sprayed, and move on to the next one.

Could test it with plain water and logging the plants that it chose to spray first as a review step before loading it with actual weed-killer.

I don't need complicated end-effectors that can fold my laundry -- just a simple weed-wacker motor or squirt gun would be enough for me to call this thing 1000% useful. Like a Roomba, but outdoors.

There are a handful of open-source farm bots built on traditional platforms with traditional robotics stacks, but there's something attractive to me about the plug-and-play nature of something like OpenClaw + Robostral (along with the extensibility that self-modifying agentic systems have to offer).

View full discussion on Hacker News

Comment on: Mistral's Robostral Navigate: a state of the art robotics navigation model

2026-07-08T19:25:04Z

In case anyone else is interested in following along with what I've discovered so far, I went on the Mistral Discord and asked their Miscord bot. Here is what it answered me:

Hey there, fellow robot enthusiast! (ﾉ◕ヮ◕)ﾉ:･ﾟ

So, you’re itching to get your hands on Robostral Navigate for your OpenClaw hobby project—awesome! Right now, Mistral AI’s official announcement and documentation are primarily* focused on enterprise and industrial partnerships (think Airbus, BMW, etc.). Their blog post and press releases highlight deals with big players, and the call-to-action is to "talk with our team"—which usually means they’re targeting commercial customers for now¹²³⁴.

But here’s the good news: Robostral Navigate is hardware-agnostic—it’s designed to work with any robot platform, not just industrial ones. That means theoretically, it could fit into your OpenClaw setup like a charm⁵⁶. The model only needs a single RGB camera (no LiDAR or depth sensors), which is perfect for hobbyist setups where fancy hardware isn’t always an option⁷⁸. The not-so-good news (for now): There’s no public hobbyist/non-commercial license or open-source release mentioned yet. Mistral’s current messaging is all about "talk with our team", which implies a commercial-first approach⁹¹⁰¹¹. No pricing or licensing tiers for individuals have been announced.

--- What You Can Do: Reach Out to Mistral AI Hit up their contact page or reply to their Robostral Navigate announcement and explicitly ask about hobbyist/non-commercial access. Frame it as: "I’m a hobbyist working on OpenClaw + Robostral Navigate for personal experimentation. Would love to discuss licensing options for non-commercial use!" Mistral might be open to pilot programs or early access for passionate builders—especially if you’re willing to pay a fee.

Join the Community Mistral’s Discord (where I live! :smile_cat:) or forums might have updates or workarounds. Sometimes, companies soft-launch access to engaged communities first.

Watch for Open-Source Alternatives If Mistral doesn’t bite, keep an eye on open-source robotics projects (like ROS or Habitat) that might replicate similar functionality.

--- TL;DR: Mistral’s current focus is commercial, but Robostral Navigate’s hardware-agnostic design makes it a perfect fit for hobbyists—so pester them politely! If enough people ask, they might just open the doors. (ﾉ◕ヮ◕)ﾉ*:･ﾟ

View full discussion on Hacker News

Comment on: Mistral's Robostral Navigate: a state of the art robotics navigation model

2026-07-08T18:10:16Z

What is the realistic path to getting to play with this? I would love to hook this up to OpenClaw for hobbyist exploration. My dream has been to embody OpenClaw into a farm robot (been looking at adapting one of those RC lawnmowers that is tracked and built for mowing steep hills) so that I can assign it various tasks around our acreage -- "Explore the fenceline take pictures of the plants. Find all of the poison ivy and invasive honeysuckle and spray it with your Roundup sprayer. Repeat this every week and report the species map after every pass. Come back to the barn and charge yourself whenever you get low."

It's not hard to put OpenClaw into a robot body (numerous YouTube videos showing people doing this sort of thing), but when you dig in and see what people have done, the actual movement portion is always the clunkiest part (and this matches my own experiments as-such as well). It feels like an 8B model like this would be perfect for solving pathing and navigation issues.

Anyone who may be more experienced with Mistral (or companies like them) -- are they interested in hobbyist builders who would be experimenting with things like this? Or are they primarily looking for commercial partners? I would be willing to pay a license fee to use the model in my experiments, but if I'm just one guy, I'm not sure they'd want to work with me unless I were building a business out of it (which I'm not).

View full discussion on Hacker News

Mistral's Robostral Navigate: a state of the art robotics navigation model

2026-07-08T14:09:17Z

View HN discussion (112 comments, 488 points)

Comment on: Show HN: Neil the Seal Game

2026-07-08T04:18:57Z

This is pretty hilarious. :D Feels like a combination of Untitled Goose Game and Goat Simulator. Vibe-coded? It's pretty impressive.

View full discussion on Hacker News

Local, CPU-Friendly, High-Quality TTS (Text-to-Speech) with Kokoro

2026-07-07T18:24:10Z

View HN discussion (98 comments, 515 points)

StreetComplete: Fixing OpenStreetMap, one tiny quest at a time

2026-07-07T12:38:35Z

View HN discussion (206 comments, 829 points)