frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

What If You Spent Every Waking Moment Taking on Elon Musk?

https://www.nytimes.com/2025/10/31/business/elon-musk-aaron-greenspan-tech.html
1•mikhael•57s ago•0 comments

YouTube denies AI was involved with odd removals of tech tutorials

https://arstechnica.com/tech-policy/2025/10/youtube-denies-ai-was-involved-with-odd-removals-of-t...
1•pseudolus•5m ago•0 comments

"Our research is greatly sped up by AI but AI still needs us"

https://twitter.com/wtgowers/status/1984340182351634571
2•wrong-mexican•10m ago•0 comments

Serious Coding: The human–AI discipline for accurate and reliable development

https://isoform.ai/blog/manual-serious-coding-vs-vibe-coding
3•Chrisywz•12m ago•0 comments

Happy Halloween, HN

6•avonmach•19m ago•2 comments

Families say cost of housing means they'll have fewer or no children

https://www.npr.org/2025/10/31/nx-s1-5551108/housing-costs-birth-rate
2•toomuchtodo•19m ago•2 comments

Show HN: DeepShot – NBA game predictor with 70% accuracy using ML and stats

https://github.com/saccofrancesco/deepshot
2•frasacco05•20m ago•0 comments

Viagrid

https://github.com/opulo-inc/viagrid
1•weinzierl•20m ago•0 comments

The goal is right. The plan is fiction

https://federalnewsnetwork.com/commentary/2025/10/the-goal-is-right-the-plan-is-fiction/
1•tailefer•20m ago•0 comments

I Love My Wife, My Wife Is Dead

https://www.bingqiangao.com/poetry/i-love-my-wife-my-wife-is-dead
2•nsoonhui•23m ago•0 comments

Large Language Models Get All the Hype, but Small Models Do the Real Work

https://www.wsj.com/tech/ai/large-language-models-get-all-the-hype-but-small-models-do-the-real-w...
1•sonabinu•26m ago•0 comments

Show HN: Unrav.io Chrome Extension – Turn any page into insights

https://unrav.io
1•rriley•26m ago•0 comments

Ask HN: Folks who had taken VEP (voluntary layoff), was it a good decision?

2•locust101•27m ago•0 comments

Grady Booch's CS Library

https://www.librarycat.org/lib/gbooch
1•mooreds•30m ago•0 comments

Livestream of Last Iceland McDonald's Burger and Fries (10 Years Old)

https://snotrahouse.com/last-mcdonalds/
1•surprisetalk•31m ago•0 comments

Wine 10.18 (Dev) – Run Windows Applications on Linux, BSD, Solaris and macOS

https://gitlab.winehq.org/wine/wine/-/releases/wine-10.18
1•neustradamus•34m ago•0 comments

Tree Tools for Schools

https://www.treetoolsforschools.org.uk/activitymenu/?cat=tree_id
1•tetris11•35m ago•0 comments

What's new in Swift: October 2025 Edition

https://www.swift.org/blog/whats-new-in-swift-october-2025/
1•frizlab•36m ago•0 comments

Little KWin Helpers

https://blog.broulik.de/2025/10/little-kwin-helpers/
1•LorenDB•36m ago•0 comments

The Hasselblad Cameras of Project Mercury

https://www.thequantumcat.space/p/the-hasselblad-cameras-of-project
1•LorenDB•36m ago•0 comments

DPRK Adopts EtherHiding: Nation-State Malware Hiding on Blockchains

https://cloud.google.com/blog/topics/threat-intelligence/dprk-adopts-etherhiding
2•hentrep•46m ago•0 comments

Show HN: Local Push-to-Transcribe

https://github.com/spacefarers/Transcrybe
2•spacefarers•47m ago•0 comments

Biotech Hunter – Crunchbase for Life Sciences

https://biotechhunter.com/
2•johnys•51m ago•0 comments

WASM-Adventure

https://github.com/euhmeuh/wasm-adventure
1•todsacerdoti•52m ago•0 comments

Can you get a star in Super Mario 64 using only one button? [video]

https://www.youtube.com/watch?v=-7VhlsqeeqI
1•unleaded•55m ago•0 comments

The Department of Defense Wants Less Proof Its Software Works

https://www.eff.org/deeplinks/2025/10/department-defense-wants-less-proof-its-software-works
5•hn_acker•58m ago•1 comments

The Situation: How Much Less Free Are We?

https://www.lawfaremedia.org/article/the-situation--how-much-less-free-are-we
2•hn_acker•59m ago•0 comments

Will Paramount Cancel Jon Stewart?

https://www.newyorker.com/culture/the-new-yorker-interview/will-paramount-cancel-jon-stewart
3•mitchbob•1h ago•2 comments

Interesting Research Programs from the 2010s (2020)

https://bcmullins.github.io/interesting-research-2010s/
2•vinhnx•1h ago•0 comments

TeraAgent: Simulating Half a Trillion Agents

https://arxiv.org/abs/2509.24063
1•jonbaer•1h ago•0 comments
Open in hackernews

From Lossy to Lossless Reasoning

https://manidoraisamy.com/lossless-reasoning.html
1•QueensGambit•6h ago

Comments

QueensGambit•6h ago
Hi HN,

OP here. In my previous post [1], I argued that code generation is the kingpin behind reasoning models. The bottleneck is LLMs generate code lossily due to tokenization fragmentation and treating code like natural language instead of structured graphs. In this post I propose:

1. Parsing user prompts into input graphs (using controlled English like ACE)

2. Parsing code into output graphs (AST trees)

3. Using graph transformers to map input graphs → output graphs

Core hypothesis: eliminating tokenization fragmentation ("ladlen" → ["lad", "len"]) and preserving tree structure could improve FrontierMath accuracy from 26% to 35-40%. No benchmarks yet. Just theory and a plan to test the improvement.

I've built compilers, not transformers, so would love technical feedback on:

- Is tokenization & linear structure really the bottleneck in code generation, or am I missing bigger issues?

- Is 35-40% improvement plausible, or overly optimistic?

- For those working on graph transformers: what approaches look promising?

Thanks in advance!

[1] Previous post - https://manidoraisamy.com/reasoning-not-ai.html

HN thread - https://news.ycombinator.com/item?id=45683113