alt.hn

5/14/2026 at 3:35:14 PM

Show HN: Watch a neural net learn to play Snake

 https://ppo.gradexp.xyz/

by c1b

5/15/2026 at 5:21:32 PM

Really cool! But right as it was nearing 4,000, it seems to have corrupted itself and no longer got any scores above 0. Not sure if that's a code bug or a neural net issue.

avg500 -4.6 last 500 episodes

peak 3959.3 best window

roll/s 20.68 20-step avg

progress 4388 562749 episodes

by ldoughty

5/15/2026 at 7:14:16 PM

Yes it just collapses eventually — never stabilizes. The training process is flawed, I suspect it has to do with the fact that some weights blow up over time, you can see in “weights” tab.

But at around 4K avg score you should see it solve the env almost every time.

Just a demo :) optimized for speed over stability.

Reward structure: Step: -1 Dot: +100 Win: +1000 so ~4k is max theoretical score on 6x6.

by c1b

5/15/2026 at 10:40:49 PM

maybe because it doesn't understand "done"? perfect play is impossible, random variance will cause scores to drop even if the model plays well and "wins". feels like it would get stuck in a loop trying to improve what can't be improved.

by ticulatedspline

5/15/2026 at 11:05:41 PM

The optimizer doesn't need to understand anything it's just an iterated mathematical construct. The author simply didn't bother to implement the necessary details to ensure numerical stability.

Alternatively it might be a problem with the scoring model in the end game.

by fc417fc802

5/17/2026 at 12:38:30 AM

That is what I thought op was saying when he used the word "understood". No need to jump on people using every day language that is still easily understood in context IMO.

by Tostino

5/15/2026 at 11:59:16 PM

feels like it would get stuck in a loop trying to improve what can't be improved.

That is the point, there is nothing on an intention that we cannot improve, the goal here is no more than 1 unique iteration of the same path

by jesuo

5/15/2026 at 6:13:01 PM

I think I noticed it reach “end game.” The snake reaches a point where, if it gets any longer, it is out of squares and hits its own tail. So it finds the route through the squares that it can infinitely loop, never eats the ball, and score starts dropping and goes negative.

by r3trohack3r

5/15/2026 at 3:58:19 PM

Cool project!

I noticed that if you go from training to watch and then back, the training temporarily drop significantly in score.

by simedw

5/15/2026 at 6:14:31 PM

It seems to be something related the moving average calculation. So it is just a glitch on the chart.

by bguberfain

5/18/2026 at 1:47:15 AM

Any plans to open-source the repo?

by jasonli0226

5/16/2026 at 1:56:48 AM

FYI this website sets off a bunch of Bitdefender alerts as being a suspicious web page. I assume probably false positives or something but still something you might want to look into.

"The page https://ppo.gradexp.xyz/ has been detected with suspicious activity. It is not recommended to continue browsing this website."

Same for:

https://ppo.gradexp.xyz/version.js

https://ppo.gradexp.xyz/dist/sizes.js

https://ppo.gradexp.xyz/dist/size_6/manifest.j

https://ppo.gradexp.xyz/dist/size_6/weights.safetens

https://ppo.gradexp.xyz/dist/sokol/demo.wa

by starshadowx2

5/16/2026 at 7:20:00 AM

it's using webgpu kernels, probably a false positive

by lucrbvi

5/16/2026 at 1:15:39 AM

Mesmerizing - could be its own digital art showcase XD Love what you've done here, friend. Looking forward to what you do next. <3

by anthonycoslett

5/15/2026 at 7:15:37 PM

did a pretty similar thing last month for the text rendering library last month.

trained and made a viz for the model and then made it displace text.

should probably do a proper write-up:https://x.com/i/status/2038367016969724259

by snats

5/15/2026 at 9:07:57 PM

I noticed snake gets penalized for not getting to the apple early, is that what you really want? Snake is about how long it gets not about the balance between length and wall clock time

by ziofill

5/15/2026 at 9:16:14 PM

But if not the snake could go into an infinite loop, never growing, never eating.

by foo12bar

5/16/2026 at 6:13:58 AM

Why? It should get the reward for getting longer, but not for getting longer quicker

by ziofill

5/16/2026 at 11:58:18 AM

Because the sessions would last forever. Think of a 1 or 2 length snake, figuring out that left down up right over and over again doesn't lose any points. You're now trapped in a local minimum. You need to make the AI get impatient (lose points) or it'll never learn.

by foo12bar

5/16/2026 at 3:52:43 PM

I see what you are saying but then wouldn’t it miss out on the best strategies, which do require patience and not going straight for the apple?

by ziofill

5/16/2026 at 7:12:52 PM

Maybe you could make it lose points for repeating a board state, I guess.

by foo12bar

5/15/2026 at 9:11:35 PM

Poorly programmed, it doesn't learn from its mistakes, the games get stuck in a loop because the snake doesn't capture a piece but the piece remains and there's a gap, constantly moving the snake along the same path with negative scores in an infinite loop leaving an unaltered yin and yang ;) there's a repetitive pattern in these infinite games between the position of the gap and the piece

by jesuo

5/15/2026 at 9:58:44 PM

Did you let it train? This doesn’t happen for me

by spectre9

5/15/2026 at 10:08:45 PM

Yes, thousands of games, you can see how it happens in the displayed game matrix, there comes a point when they all enter those loops https://ibb.co/bM4RPzPb

by jesuo

5/15/2026 at 10:36:16 PM

Makes sense, author mentioned training collapses eventually

by spectre9

5/16/2026 at 7:47:14 PM

This is seriously impressive. Running PPO training directly in the browser through WebGPU feels like a glimpse into where lightweight AI experimentation is headed.

by Mariajaved906

5/15/2026 at 4:37:41 PM

My average eventually made it to about 3900, and then stagnated between 3600-3900. I'm curious if this is universal behavior or not. I'm up to about 5k steps.

by beardsciences

5/16/2026 at 8:57:45 AM

Give the neural network the sense of sight, to know where the point is located.

by jesuo

5/15/2026 at 4:06:37 PM

More details and implementation notes please?

by neduma

5/15/2026 at 8:08:19 PM

It's on the page, if you click the little info icon in the upper-right. Here's the text but there's some nice graphics there too:

  Snake Game, training entirely in the browser. Built on tinygrad: the rollout / targets / train graphs are TinyJits authored in Python, then compiled once to WGSL and replayed here under WebGPU.

  Observation: flat 10×10 board (100) + 4-dim prev-action one-hot = 104 dims. fc_pi.weight is zero-init so the opening policy is uniform over the legal actions; fc_v uses tinygrad's default Kaiming init.

  Per rollout: T=24 × N=384 parallel snakes (9,216 transitions), then K=3 epochs × 4 mini-batches of PPO updates. GAE γ=0.99, λ=0.95; AdamW wd=0.01; ratio clip ε=0.1; grad-norm 0.5; Huber value β=1, val_coef=1; entropy bonus 0.008333333333333333.

  Action mask + value clip + KL early stop. The 4-dim prev_a obs tail lets fc_pi zero the U-turn logit (the env silently overrides same-axis reversals anyway). Value loss is max(huber(v_new−td), huber(v_clip−td)) at ε=0.2. Approx-KL is sampled after each epoch and breaks the loop at 1.5·kl_target.

by cshimmin

5/15/2026 at 5:42:15 PM

> WebGPU not available in this browser

Looks like this is for Linux and Windows, on NetBSD I get this issue :(

by jmclnx

5/15/2026 at 6:39:14 PM

I got this in Firefox on Linux, just had to enable WebGPU in about:config (`dom.webgpu.enabled` = true).

by NoboruWataya

5/15/2026 at 6:56:42 PM

Did not know that existed, I enabled it but no luck. Must be a NetBSD thing based upon this new message:

> WebGPU is not yet available in Release or late Beta builds.

by jmclnx

5/15/2026 at 6:09:44 PM

If you are using brave (which i assume also applies to chrome) , there is a menu at brave://flags , you can enable unsafe web GPU from there

by redshiftza

5/15/2026 at 8:13:16 PM

That's cool, i did exactly the same few years ago

by mavdol04

5/16/2026 at 12:08:50 AM

sound cool; would like to show my kid for education; doesn't work on Mac/Safari though (no webGPU)

by insane_dreamer

5/16/2026 at 4:32:10 AM

You can enable it in settings; works on my older iPhone.

by c1b

5/16/2026 at 12:02:39 PM

Very cool! Not GitHub repo?

by dennisy

5/15/2026 at 6:25:06 PM

Link to repo?

by LowLevelKernel

5/16/2026 at 10:31:34 AM

damn this was really interesting and really well executed

by ionwake

5/16/2026 at 4:40:13 AM

Will be open-sourced?

by modinfo

5/15/2026 at 5:08:49 PM

cool project

by th1nhng0

5/15/2026 at 7:02:31 PM

Crashed

by bozhark

5/17/2026 at 1:14:31 PM

[dead]

by RyanJohn

5/16/2026 at 1:30:07 PM

[dead]

by coder0x

5/15/2026 at 5:11:30 PM

my training on a 10x10 just randomly broke. i got to like 3600 then the graph went flat, the viewer on the left just showed it constantly restarting the game, and the scores in the negative. my average is now -10.

by austinthetaco

5/16/2026 at 9:56:31 AM

[flagged]

by Ozzie-D

5/16/2026 at 8:09:43 PM

[dead]

by NexiunDev