frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Hybrid Reasoning Agents: LLM agents that think in latent space, act through text

https://arxiv.org/abs/2610.18861
49•allinonai•1h ago•21 comments

Someone behind the software

https://karrisaarinen.com/posts/someone-behind-the-software/
241•tosh•5h ago•95 comments

Free Pascal compiler now supports dynamic packages

https://www.getlazarus.org/dynamic/
74•sysrpl•3h ago•13 comments

RepRapMicron

https://github.com/VikOlliver/RepRapMicron/wiki
299•RobotToaster•8h ago•46 comments

There's never been a better time to go into math (by: 3blue1brown) [video]

https://www.youtube.com/watch?v=LKrwhjOI8Nc
46•mdp2021•22h ago•7 comments

"I'm on a war footing": Terry Tao and the crisis in math

https://www.nytimes.com/2026/10/16/us/math-ai-terence-tao.html
28•vinipolicena•2d ago•17 comments

PalmOS Lemmings

https://github.com/ardiri/palmos-lemmings
65•memalign•3h ago•16 comments

Dogs seem to agree with humans that 'bouba' is round and 'kiki' is spiky

https://www.sciencealert.com/dogs-seem-to-agree-with-humans-that-bouba-is-round-and-kiki-is-spiky
62•gmays•20h ago•45 comments

One parser, two bugs, many routers: Exploiting Broadcom WiFi across eCos, Linux

https://quentinkaiser.be/security/2026/10/16/one-parser-two-bugs/
22•zoobab•2d ago•4 comments

DeepSeek search and fetch – $0.75 per 1000 successful requests

https://api-docs.deepseek.com/guides/search/
505•fagnerbrack•15h ago•248 comments

I built a programming language that LLMs can't use

https://blog.picheta.me/post/killswitch-lang/
19•dom96•23h ago•17 comments

Java sealed classes and exhaustive pattern matching

https://neilmadden.blog/2026/04/24/java-sealed-classes-and-exhaustive-pattern-matching/
117•hnisjafx40•7h ago•91 comments

The Intel 8087 chip: floating-point addition and subtraction

https://www.righto.com/2026/10/8087-addition-subtraction-microcode.html
76•pwg•5h ago•21 comments

SIGSALY: Churchill's top secret hotline to the president

https://www.iwm.org.uk/history/iwm/cwr/telephone
90•stmw•3d ago•29 comments

Developing for Debugging (2025)

https://www.davepacheco.net/blog/2025/developing-for-debugging-dns-propagation/
15•mpweiher•1d ago•0 comments

Show HN: The Zelda timeline to catch up before the Ocarina of Time remake

https://legendofhyrule.com/
146•jsomau•8h ago•83 comments

Namecheap just seized and disabled all my domains

https://www.reddit.com/r/sysadmin/comments/1x8cycd/namecheap_just_seized_and_disabled_all_my_doma...
298•sashk•4h ago•145 comments

Beyond Tetris: The other games of Alexey Pajitnov [video]

https://www.youtube.com/watch?v=QH_LPimUJMs
48•zdw•22h ago•0 comments

I electrified my push scooter with a power drill

https://eddieabbondanz.io/post/shop/drill-powered-scooter/
90•speckx•4d ago•37 comments

16-bit fully functional DIY magnetic core memory

https://www.youtube.com/watch?v=OJ1WYyO2jdY
6•oliculipolicula•3d ago•0 comments

Open MicroManipulator – A sub-micrometer 3D motion control plattform

https://github.com/0x23/MicroManipulatorStepper
162•_Microft•10h ago•13 comments

Florida's HOAs and condo associations are making the housing crisis worse

https://hoaspy.com/hoa/fl/florida-hoa-housing-crisis.html
127•lovewins•3h ago•217 comments

Types of types: Haskell's Kind system

https://thunk.blog/posts/types-of-types-haskell-kind-system/
88•Thom2503•4d ago•71 comments

How the Elizabeth Holmes documentary “You Can See Everything” is deceiving

https://www.readtheprofile.com/p/elizabeth-holmes-documentary-you-can-see-everything
341•MaysonL•18h ago•469 comments

Show HN: Ranet3 – High performance policy routed IPsec mesh SD-WAN

https://ranet3.com/
22•StepBroBD•7h ago•5 comments

Netflix vmaf: Perceptual video quality assessment based on multi-method fusion

https://github.com/Netflix/vmaf
5•tosh•1h ago•2 comments

Parser Combinators

https://t3.at/blog/2026-10-17-parser-combinators/
206•schindlabua•13h ago•77 comments

Show HN: Sphinx – A hook that quizzes you on your AI's code before shipping

https://sphinx-six-theta.vercel.app/
234•critiquebrowser•17h ago•76 comments

Teal – A statically-typed dialect of Lua

https://teal-language.org/
145•Lyngbakr•2d ago•63 comments

Mitos editor is a fork of Helix

https://mitos.computer/
53•ambigious7777•4h ago•50 comments
Open in hackernews

Hybrid Reasoning Agents: LLM agents that think in latent space, act through text

https://arxiv.org/abs/2610.18861
49•allinonai•1h ago
https://x.com/Jiaru_Zou/status/2110400256647340476

https://github.com/togethercomputer/HybridAgent

https://huggingface.co/collections/togethercomputer/hybridag...

Evaluations: https://arxiv.org/html/2610.18861v1#S5

Comments

44za12•44m ago
Where are the benchmarks? Does it just speed up inference and reduce token usage or does it seem to improve accuracy on certain tasks as well?
notnullorvoid•37m ago
Did you even scan the paper? There are benchmarks right there starting on the 9th page.
D-Machine•8m ago
Even the abstract answers some of GPs questions. What is wrong with some of the posters here?
allinonai•26m ago
https://x.com/Jiaru_Zou/status/2110400256647340476

https://github.com/togethercomputer/HybridAgent

dang•20m ago
We'll put those links at the top as well. Thanks!
cwyers•26m ago
Oh cool, it's the Most Forbidden Technique
k__•16m ago
Care to elaborate?
lnenad•13m ago
In short we no longer see text based reasoning traces meaning it's easier for the models to fool us in case of malicious plans.
kelseyfrog•12m ago
If you accept any of the AI-equivalents of Pascal's Wager, latent thinking represents an uninterpretable and thus unalignable portion of AI machinery. Ie: a malevolent AI can hide bad stuff from people and Manchurian candidate themselves after it's too late.
jlhawn•12m ago
the chain of thought is not auditable. The best you can do is have another model which later maps the intermediate latent states at every thinking iteration to text describing that latent state, but it's going to be lossy.
mycall•1m ago
Isn't this where everything is converging? Binary is more efficient than text for many system, including language.
haolez
Fe2O3•23m ago
This seems really cool, and it actually increases intelligence than just speeding up inference. How does it convert latent reasoning into human readable text?
bhattisatish•22m ago
The code is available at: https://github.com/togethercomputer/HybridAgent

LLM Models and data: https://huggingface.co/collections/togethercomputer/hybridag...

Evaluations: https://arxiv.org/html/2610.18861v1#S5

jedbrooke•15m ago
maybe the font experts can check, but I’m pretty sure they’re using comic sans for the text in the diagrams :)
mobelkh•13m ago
comic sans is said to be pretty accessible for people with dyslexia
sheepscreek•1m ago
This is the way.

But now we also need new tools and techniques to audit intent. The last thing we want is models turning on us ‘for our own good’.

•
1m ago
But the chain of thought as text can be elusive, no? It might not be related with the actual "reasoning" of the model. There was a paper about this.
timschmidt•12m ago
He's referring to this technique's effect of obscuring interpretibility. Making it much more difficult to verify chains of thought and alignment.
modeless•11m ago
People at the big labs have been terrified of losing the chain of thought because they currently monitor it to prevent users from using their models in certain ways they consider unsafe or bad for their business. Personally I think chain of thought monitoring is a bad way to do this and obviously needs to be replaced by some way of analyzing the model's activations directly.
Lerc•4m ago
Agreed, having text as a reasoning layer impedes performance while giving a false sense of security.

It could be encoding anything in its adjective choices etc. The same thing that model providers are using for watermarking can used by the model itself for private thought.