frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

How many e's are in the word seventeen [video] (AI hallucination)

https://www.youtube.com/shorts/nks72LuZO20
6•Imustaskforhelp•1h ago

Comments

Imustaskforhelp•1h ago
I really found this fascinating as I had thought that these type of problems of how many e's are in the word strawberry etc. were stopped but as this video shows, perhaps its just that this question of how many e's are in the word strawberry itself got part in the training data and so even a slight variation of asking it for seventeen makes it fumble. I had thought that this was a solved issue but actually it isn't which was a bit fascinating to see in this video, so much so that I had to test it out and I found out that AI still hallucinates and had the same result for the most part.
CamperBob2•12m ago
The Qwen 3.6 27B 8-bit quant has no problem with it. I'd guess that most thinking models won't fail this kind of test anymore, while some base or instruct models that are not post-trained for reasoning will still fail it.

I also can't reproduce it in ChatGPT 5.3 Instant with auto-thinking disabled. Solved problem, as far as I'm concerned. Maybe this particular case was a bug in the voice model, or just some BS the YouTuber made up for clicks. (Notice that we never actually see the answer in text form.) Mission accomplished, I guess.

wvbdmp•1m ago
There is an old German joke from the comedy “Die Feuerzangenbowle”, where someone gives his name as “Pfeiffer, with 3 fs”, which is funny because it is robotically hypercorrect, since everybody knows the only necessary clarification is between “Pfeiffer” vs. “Pfeifer”, double-f vs. single f. So in a way “there are two rs in strawberry” is a very human “mistake”, because in any normal situation the asker is clearly interested in the “berry” part. This weird sycophancy, however, is entirely preposterous and hopefully just an artifact of some deliberate “the customer is always right” policy corporate tacked on, rather than a fundamental limitation of the technology.

Dayton mayor demands accountability after plate-reader data breach

https://www.wdtn.com/news/mayor-commissioner-demand-alpr-data-breach-accountability/
1•ok123456•48s ago•0 comments

Show HN: Cuqueclicker (Cookieclicker Inspired Game)

https://flipbit03.github.io/cuqueclicker/
1•fb03•2m ago•0 comments

Why TUIs Are Back

https://wiki.alcidesfonseca.com/blog/why-tuis-are-back/
1•rickcarlino•2m ago•0 comments

Safe(R) Repo Access for Agents

https://obiwahn.org/posts/safe-sftp-access-for-agents/
1•obiwahn•4m ago•1 comments

Cheap worktree replacement for agent swarm

https://github.com/satmihir/wafers
1•0state•5m ago•0 comments

Writing the loss function: AI, feeds, and the engagement optimizer

https://eignex.com/posts/writing-the-loss-function/
1•monom•6m ago•1 comments

PDF Editify Merge PDF Tool

https://pdfeditify.com/merge-pdf
1•Benish_Zulfiqar•7m ago•0 comments

Gratitude for an Open Android

https://lzon.ca/posts/series/grateful/keep-android-open/
1•jpmitchell•9m ago•0 comments

I Use Codex CLI to Write and Maintain a Book on Codex CLI

https://blog.danielvaughan.com/how-i-use-codex-cli-to-write-and-maintain-a-book-on-codex-cli-0480...
1•dvaughan•13m ago•0 comments

The Audio Industry Is Grappling with the Rise of 'Podslop'

https://www.bloomberg.com/news/newsletters/2026-04-30/-podslop-proliferation-is-challenging-the-a...
2•thm•15m ago•0 comments

Brush a Bash/POSIX-compatible shell written in the Rust

https://github.com/reubeno/brush/releases/tag/brush-shell-v0.4.0
2•Jarlakxen•15m ago•0 comments

The Death of Scrum – Built for a slower world, performed by those who left

https://death-of-scrum.net/
2•mantyx•16m ago•0 comments

The Croupier – AI-powered candidate discovery and ranking for recruiting teams

https://dealers-choice-systems.vercel.app
1•Berlia•16m ago•0 comments

Camera Obscura #1

https://www.maartenheijkamp.nl/portfolio_page/camera-obscura-1/
1•jruohonen•16m ago•0 comments

Fake Money Saved Brazil (2010)

https://www.npr.org/sections/money/2010/10/04/130329523/how-fake-money-saved-brazil
1•downbad_•19m ago•1 comments

Ferrari Luce

https://www.ferrari.com/en-EN/auto/ferrari-luce
1•varun_chopra•19m ago•1 comments

Strategic Orbit Wars competition on Kaggle

https://www.kaggle.com/competitions/orbit-wars
1•darshanmakwana•19m ago•0 comments

Data Breach Hits Anti-ICE Organizing Site Gtfoice.org

https://substack.com/home/post/p-196263468
2•CapricornNoble•20m ago•0 comments

Wheel Reinvention: Not Just a Common Metaphor

https://tedium.co/2026/05/03/wheel-reinvention-technology-history/
1•shortformblog•21m ago•0 comments

Talking to Transformers

https://miraos.org/blog/2026/05/02/talking-to-transformers
2•taylorsatula•27m ago•0 comments

Supercollider

https://supercollider.github.io/
2•andyjohnson0•29m ago•0 comments

Separating what AI does well from what code does well

https://claude.com/blog/how-kepler-built-verifiable-ai-for-financial-services-with-claude
2•eddiehammond•29m ago•1 comments

Original Volkov Commander Sources

https://github.com/ddanila/vc
2•begoon•31m ago•0 comments

Mapping Notable Geo-Triangles

https://nickscip.xyz/projects/triangles/
2•nickscip•31m ago•1 comments

Open source desktop app for studying online courses

https://github.com/tonhowtf/omniget
2•axiomdata316•33m ago•0 comments

Show HN: Security Scanner for Agent Skills and MCP

https://github.com/snyk/agent-scan
3•lirantal•34m ago•0 comments

Lyra 2.0: Explorable Generative 3D Worlds

https://research.nvidia.com/labs/sil/projects/lyra2/
1•jonbaer•35m ago•1 comments

Past Ferrari Models, 1947–2023

https://www.ferrari.com/en-US/auto/past-model
1•NaOH•35m ago•0 comments

Show HN: A Privacy tool that finds and hides sensitive data in phtots/videos

https://www.scanon.ai/
1•lotuslabs•35m ago•0 comments

A GitHub for Maintainers

https://nesbitt.io/2026/05/02/a-github-for-maintainers.html
1•milkglass•35m ago•1 comments