frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

Shutting Down Clear Linux OS

https://community.clearlinux.org/t/all-good-things-come-to-an-end-shutting-down-clear-linux-os/10716
2•todsacerdoti•2m ago•0 comments

Nuxt Joins Vercel

https://vercel.com/blog/nuxtlabs-joins-vercel
1•rattray•7m ago•1 comments

The Kap Programming Language

https://kapdemo.dhsdevelopments.com/examples.html
2•thunderbong•13m ago•0 comments

A Software for One

https://www.jasonthorsness.com/30
2•jasonthorsness•13m ago•0 comments

Women Are Falling Behind in America's Return to the Office

https://www.wsj.com/lifestyle/careers/return-to-office-gender-gap-236392aa
4•bdev12345•14m ago•0 comments

Astronomer launches internal investigation after viral Coldplay video

https://www.cnn.com/2025/07/18/entertainment/coldplay-concert-kiss-cam-astronomer-investigation
2•bb88•14m ago•0 comments

Build your CV on Subreply as a LinkedIn alternative

https://subreply.com/lm
4•lcnmrn•18m ago•0 comments

Curse Not the King

https://daringfireball.net/2025/07/curse_not_the_king_cbs_colbert_trump
2•Bogdanp•18m ago•0 comments

The Physics of Dissonance (MinutePhysics) [video]

https://www.youtube.com/watch?v=tCsl6ZcY9ag
1•jerf•23m ago•0 comments

Billionaire Gabe Newell: pitching VCs makes no business sense

https://www.pcgamer.com/gaming-industry/multi-billionaire-gabe-newell-says-the-whole-startup-culture-of-pitching-vcs-for-capital-makes-no-business-sense-a-great-way-of-destroying-money-and-wasting-peoples-time/
6•e2e4•24m ago•0 comments

Ccusage: A CLI tool for analyzing Claude Code usage from local JSONL files

https://github.com/ryoppippi/ccusage
9•kristianp•25m ago•2 comments

Fuzzing macOS Userland (For Fun and Pain)

https://marqcodes.com/fuzzyingforfun.html
1•N3Xxus_6•26m ago•0 comments

Free Online Minesweeper

https://www.freeonlineminesweeper.com
1•avonmach•26m ago•0 comments

DHH – I Hate TypeScript (3 min video)

https://www.youtube.com/watch?v=tyjUH5TLSTM
3•rmason•32m ago•0 comments

Show HN: Interactive Bash tutorial that runs in the browser

https://sandbox.bio/tutorials/bash-script
2•raboukhalil•35m ago•0 comments

Show HN: Castream – Native iOS/Android IRL multistreaming app

1•acabralto•35m ago•0 comments

There Is No Antimemetics Division – A Novel (2025)

https://www.penguinrandomhouse.com/books/783041/there-is-no-antimemetics-division-by-qntm/
2•Duanemclemore•39m ago•1 comments

First earthquake, then fire: UC San Diego researchers test steel building

https://www.kpbs.org/news/science-technology/2025/07/17/first-earthquake-then-fire-uc-san-diego-researchers-test-steel-building
2•littlexsparkee•41m ago•1 comments

Ask HN: What are your favorite open source AI agent implementations?

2•kanodiaashu•41m ago•0 comments

Node.js 18 is being deprecated

https://vercel.com/changelog/node-js-18-is-being-deprecated
1•ananddtyagi•46m ago•0 comments

EPA says it will eliminate its scientific reseach arm

https://www.nytimes.com/2025/07/18/climate/epa-firings-scientific-research.html
36•anigbrowl•47m ago•4 comments

Vibe coding? AI assisted coding? I prefer being an AI micromanager [video]

https://www.youtube.com/watch?v=3gnfOnhC1EA
5•godot•52m ago•0 comments

"Pitch in " Anti-Litter PSA (1973) [video]

https://www.youtube.com/watch?v=Sba0GzhZ088
1•petethomas•56m ago•0 comments

Agents Built from Alloys

https://xbow.com/blog/alloy-agents/
2•azhenley•57m ago•0 comments

US EPA cutting workforce by 23%, closing research division

https://www.reuters.com/legal/government/us-epa-cutting-workforce-by-23-closing-research-division-2025-07-18/
15•pseudolus•1h ago•1 comments

I'm Rebelling Against the Algorithm

https://varunraghu.com/im-rebelling-against-the-algorithm/
3•Varun08•1h ago•0 comments

My worst tech purchase became my best DIY desk lamp

https://medium.com/@philwornath/when-2-useless-items-unite-repurpose-your-monitor-lamp-bar-ikeahackers-upcycling-02e6ad595e1b
1•philjw•1h ago•1 comments

Show HN: Vizr – Ask questions about your marketing data, get real answers

https://vizr.app/
1•arifliftos•1h ago•0 comments

With One Call, Trump Alters the Fate of a Contested Power Project

https://www.nytimes.com/2025/07/17/climate/hawley-grain-belt-express-invenergy-trump.html
4•zekrioca•1h ago•2 comments

Is Translation the Killer App?

https://substack.com/home/post/p-168658235
1•mathattack•1h ago•0 comments
Open in hackernews

Can LLMs Do Accounting?

https://accounting.penrose.com/
5•yunyu•4h ago

Comments

yunyu•4h ago
LLMs are on the verge of replacing data scientists and investment bankers. But can they perform simple accounting tasks for a real business?

We built AccountingBench, a test where LLMs must "close the books" for a real SaaS business using 1 year of Stripe/Ramp/Rippling/Mercury data.

Claude 4 and Grok 4 start strong - within 1% of human CPA baselines in month 1.

But as time progresses, all models inevitably accumulate compounding errors and exhibit erratic behavior, causing significant deviations.

That said, the early accuracy here is promising. With targeted post-training, models may be able to replace humans for this kind of work.

simmerup•4h ago
Accounting isn't really the type of thing that can accept errors though is it?

Like it needs to be 0% error rate

yunyu•7m ago
A certain level of errors is tolerable/inevitable. But the accountants need to be able to correct for them once they build up
bell-cot•4h ago
Given their inclination to fabricate user-pleasing answers...could I let an LLM do my tax returns?
yunyu•1h ago
No comment, the good news is that accounting and taxes are verifiable - so in principle it is possible to RL models to do them correctly
mmarian•4h ago
I was just thinking of that earlier today, really cool!
AlSweigart•4h ago
LLMs are really not good at following specific processes like math. They operate off vibes.

Ask Claude to multiply two ten-digit numbers. It gets the first one or two digits correct, and then makes up the rest.

ChatGPT used to have the same problem, but now it writes a program to perform the math for it.

yunyu•1h ago
This was true up until they started training them using Reinforcement Learning from Verifier Feedback (started with O1). By sticking a calculator in the training loop, they seem to have gotten out of the arithmetic error regime. That said, the ChatGPT default is 4o which is still susceptible to these issues.