frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Tikos – Turn any study material into a question bank (local-first, open-source)

https://github.com/Dawnfz-Lenfeng/tikos
1•Dawnfz•49s ago•0 comments

The Death of Open Channels

https://doerpmund.com/musings/death-of-open-channels
1•felixdoerp•2m ago•0 comments

Precursor

https://blog.cloudflare.com/introducing-precursor/
1•AznHisoka•3m ago•0 comments

Indian scientists are mapping the brain's last frontier

https://www.bbc.com/news/articles/cg53l737v1qo
1•akbarnama•5m ago•0 comments

Sydney MP pushes to bring human composting to Australia as burial costs rise

https://www.abc.net.au/news/2026-06-25/sydney-mp-alex-greenwich-human-composting-bill-explained/1...
1•speckx•5m ago•0 comments

LinkedIn, a mass grave of ghost jobs, is now becoming a dating app

https://www.sfgate.com/tech/article/linkedin-dating-app-22340622.php
3•bookofjoe•5m ago•2 comments

Red Hat will support your RHEL forever now – for a price

https://www.zdnet.com/article/red-hat-enterprise-linux-forever-support/
2•CrankyBear•6m ago•0 comments

There's no way this bond market can fund the markets needs without higher yields

https://www.youtube.com/watch?v=W-Dxpp9YG3s
2•root-parent•6m ago•0 comments

Detained by settlers, US Democrat Ro Khanna now faces pro-Israel attacks

https://www.aljazeera.com/news/2026/7/12/detained-by-settlers-us-democrat-ro-khanna-now-faces-pro...
2•speckx•9m ago•0 comments

Tell HN: Staged NPM publishing is awful

1•electrovir•10m ago•0 comments

AI agents 136.5 times less efficient than conventional AI

https://www.theengineer.co.uk/content/news/ai-agents-1365-times-less-efficient-than-conventional-ai
2•antondd•10m ago•0 comments

From Muon to Gradient Clipping: Some Thoughts on QK Stability

https://MasterGodzilla.github.io/posts/2025/07/muon-clip/
1•Eridanus2•10m ago•0 comments

Is Proof-of-Stake the Green Future of Crypto?

https://www.disruptionbanking.com/2026/07/13/is-proof-of-stake-the-green-future-of-crypto/
1•emsidisii•10m ago•0 comments

The Fishing Contest

https://www.quarter--mile.com/The-Fishing-Contest
2•surprisetalk•12m ago•0 comments

10 Languages in 3 days: field notes from localizing an AI translation product

https://transept.ai/journal/localization-lessons-ten-languages
1•sithamet•12m ago•0 comments

Optional standards dont cut it

https://nt-ai-dc.info
1•justatdotin•13m ago•0 comments

Prefect Is Acquiring Dagster

https://dagster.io/blog/prefect-is-acquiring-dagster
1•pbronez•13m ago•1 comments

New method aims to keep kids safe from illegal AI-generated content

https://news.mit.edu/2026/new-method-keeps-kids-safe-from-illegal-ai-generated-content-0713
1•droidjj•13m ago•0 comments

A Jupiter-size planet that escaped its star's death

https://arstechnica.com/science/2026/07/a-jupiter-size-planet-that-escaped-its-stars-death/
2•rbanffy•13m ago•0 comments

Don't ask what you want. Ask who you want to be

https://softwaredoug.com/blog/2026/07/13/who-want-to-be
1•softwaredoug•14m ago•0 comments

Show HN: Replaces Supabase and FastAPI

https://github.com/backant-io/jerrycan
1•phegler•14m ago•0 comments

Coding Is a Right

https://www.nytimes.com/2026/07/13/opinion/ai-code-free-speech.html
1•runesoerensen•15m ago•1 comments

Show HN: I made another search for the HN Who Is Hiring threads

1•devzl•15m ago•0 comments

Drydock – Agile/TDD Spec Driven Delivery – Initial Release

https://github.com/webcloudstudio/Drydock
1•happyed•15m ago•0 comments

How are people securing their AI access to APIs?

https://requestrocket.com/blog/ai-agent-access-control-landscape
1•geoicons•15m ago•0 comments

You are not in the race against slop cannons

https://hils.substack.com/p/you-are-not-in-the-race-against-slop
1•herbertl•17m ago•0 comments

Easy to USE online PDF Editor with OCR, merge PDFs

https://pixoate.com/pdf-editor
1•HSK11•18m ago•1 comments

Richard Sutton Breaks Away from Keen AI to Start Oak Lab

https://twitter.com/RichardSSutton/status/2076663628301058329
1•tosh•18m ago•0 comments

Reaction: Daemon Scanning Program Outputs for Repeated Patterns and Actions

https://framagit.org/ppom/reaction
1•birdculture•19m ago•0 comments

Truth about battery fires – Gore Street Capital [video]

https://www.youtube.com/watch?v=LDMs7QQqQxE
1•leonidasrup•19m ago•0 comments
Open in hackernews

I had rival LLMs judge 205 of my own startup ideas

https://skeptral.com
1•mdiske•52m ago

Comments

mdiske•52m ago
Hey there all, I'm a solo founder and this is my first project, so be gentle with it.

So, for the past year I did build some of my ideas I had in the past but never had enough time and today's AI helps accelerate projects's development like never before. My Major is SWE so like many of you in here maybe I did pursue some ideas just to realise that the product has been already built by a random company 2 days before, it could be killed by another update coming form the big LLM vendors or it's just not feasible.I know it sounds dumb but just getting really excited by building an idea you've had for a long time can just make you blind.

Based on my past 4 failed projects that were killed right before launch, I did create initially a system that goes all over the internet to look at people's pain points to see if my products would be targeting people's needs or will it be another product that will never see the light of the day. It got iterated a few times to actually give pain scores on every idea that came in from my ideas notebook and I did test it and refine it to actually be as merciless as possible and seek valid reason why an idea should be build or not, this has been done on a daily basis for the past months to not create a roleplay ideas roaster that can be vibe coded in less than a month.

This based truth detector led me to launch today Skeptral: - it consists of independent LLMs from different vendors and each scores an idea 0-100 - it returns Scale, Pivot, or Kill, with a sourced reason for the verdict that you can click through.

I fed it 205 of my own ideas. 41 came back killed outright, 162 wounded, and 2 survived intact. That's the dataset this post links to. You can sort it, read the kill-shots, and find the cases where the models split, which are the interesting ones.

To check it isn't just cynical, I ran a calibration set: - five deliberately absurd ideas (it killed all five) and five companies that later got huge, each worded the way it sounded before it worked - It killed Uber, Dropbox, Twitter and Stripe. - Only Airbnb survived, to Pivot.

So yes, it would have told me not to build Stripe.

I'd rather show that than hide it: - it's calibrated against absurdity, not clairvoyant about the future and the kill-shot is a checkable argument not a verdict from God.

How a split resolves: - I don't average the scores, that launders the signal. - A kill is sticky. If any model lands a sourced, obvious kill, the verdict trends Kill unless the others can rebut the source because a real reason to die beats two vibes to live.

Pasted ideas aren't stored, pooled or trained on.

There are zero testimonials, on purpose, because the whole point is no flattery.

The dataset is the proof instead.

Paste your own idea in the free box no login and try to make it survive.

If a kill-shot is wrong, reply and tell me exactly where it fails, that would be really useful!

Thank you for taking the time to read it and I hope it will help you as it did me!