frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

A GPT-5.4 bug led to OpenAI banning goblins and raccoons

5•prabal97•1h ago
Someone found this in OpenAI Codex’s system prompt:

"Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user’s query."

Goblins, gremlins, trolls, ogres - the fantasy quartet, fine. I get it. Perhaps someone asked Codex to write something and got back a Runescape goblin lore deep-dive.

But pigeons? Raccoons? These are deployment-environment animals.

If not for raccoons, who’d do the garbage collection?

RFC 1149 (IP over Avian Carriers, 1990, an actual protocol - with an actual implementation, in Bergen) puts pigeons squarely inside the systems-engineering canon.

“Or other animals or creatures” - full anti-creature agenda. Literally a bajillion dollars get set on fire every hour processing these prompts.

Turns out there’s a reason for the ban. Back in 5.4, GPT developed a goblin obsession and there are reddit threads full of users whose ChatGPT couldn’t stop bringing up gremlins. The system prompt is the exorcism that shipped with 5.5.

If you think I’m lying, see for yourselves: the prompt is right there in the repo link shared below.

REPO LINK: https://github.com/openai/codex/blob/de2ccf94735a3d8a2a7077e6a5292026413867cf/codex-rs/models-manager/models.json#L55

HN LINK: https://news.ycombinator.com/item?id=47319285

REDDIT THREAD 1: https://www.reddit.com/r/ChatGPT/comments/1roci77/anyone_elses_chatgpt_obsessed_with_goblins_since/

REDDIT THREAD 2: https://www.reddit.com/r/ChatGPT/comments/1rll8hb/suddenly_obsessed_with_goblins_and_gremlins/

Meta found in breach of EU law for failing to keep children off platforms

https://www.theguardian.com/technology/2026/apr/29/meta-found-in-breach-of-eu-law-for-failing-to-...
2•geox•2m ago•0 comments

RSME: A Reactive Stability Mutation Encryption

https://zenodo.org/records/19712564
1•RanggaS•2m ago•0 comments

Study Finds a Third of New Websites Are AI-Generated

https://www.404media.co/study-finds-a-third-of-new-websites-are-ai-generated/
1•thm•5m ago•0 comments

Who's on call? How Opus 4.6 helped us calculate this 2,500x faster

https://incident.io/blog/whos-on-call-how-claude-helped-us-calculate-this-2-500-x-faster
1•boryrain•8m ago•0 comments

Show HN: TiGrIS, a tiling compiler that fits ML models onto embedded devices

https://github.com/raws-labs/tigris
2•asteinh•13m ago•0 comments

ANP – A binary protocol for AI agent-to-agent price negotiation (no LLM tokens)

https://github.com/victornominista/anp
1•VC83•16m ago•0 comments

Rebuilding the Data Stack for AI

https://www.technologyreview.com/2026/04/27/1136322/rebuilding-the-data-stack-for-ai/
1•joozio•21m ago•0 comments

Sovereign Tech Standards network: financial support for open source maintainers

https://www.sovereign.tech/programs/standards
4•mgajdo•21m ago•1 comments

Anthropic's Champion Kit for engineers pushing Claude Code at their company

https://code.claude.com/docs/en/champion-kit
1•ashadh•21m ago•0 comments

India's major airlines on 'verge of closing down' as high fuel costs sting

https://www.scmp.com/news/asia/south-asia/article/3351782/indias-major-airlines-verge-closing-dow...
2•TMWNN•24m ago•0 comments

A Podcast with Talkie, a 13B model trained only on pre-1931 text

https://the-coming-age-aqx87j.jellypod.com
1•bilater•26m ago•1 comments

The UK's Answer to DARPA Wants to Rewire the Human Brain

https://www.wired.com/story/kathleen-fisher-jacques-carolan-aria-wired-health/
2•beardyw•31m ago•0 comments

Specification-Driven Development framework for agent-native development

https://specdd.ai
1•addvilz•34m ago•0 comments

Low-Compilation-Cost Register Allocation in LLVM-Based Binary Translation

https://dl.acm.org/doi/abs/10.1145/3767295.3803591
2•matt_d•36m ago•0 comments

Germany Overtakes US in Ammunition Production Capacity

https://www.newsweek.com/germany-overtakes-us-in-ammunition-production-capacity-11886409
13•vrganj•38m ago•2 comments

Ask HN: What are you doing during inference?

4•petesergeant•38m ago•2 comments

Scaleway raises cloud prices after 3 years of absorbing costs

https://www.scaleway.com/en/blog/a-transparent-update-on-scaleway-pricing/
1•latentframe•39m ago•0 comments

What's next for voice dictation apps

1•priyesh2000•39m ago•0 comments

PEP 661: Sentinel values, has been accepted 5 years later

https://peps.python.org/pep-0661/
2•azhenley•39m ago•0 comments

Webxdc: Secure mini apps shared in a chat

https://webxdc.org/
1•janandonly•39m ago•0 comments

Ask HN: How do you differentiate with AI coding interviews?

1•marcell•41m ago•0 comments

Using Codex and ForgeCAD to Make a Model of the Teenage Engineering KOII

https://twitter.com/theopuslabs/status/2049195007404380244
1•opuslabs•42m ago•0 comments

Virtualisation on Apple Silicon Macs is different

https://eclecticlight.co/2026/04/29/virtualisation-on-apple-silicon-macs-is-different/
2•ingve•44m ago•0 comments

AWS launches Amazon Quick desktop AI assistant

https://aws.amazon.com/quick/
1•nine_minutes•44m ago•1 comments

HardenedBSD Is Now Officially on Radicle

https://hardenedbsd.org/article/shawn-webb/2026-04-26/hardenedbsd-officially-radicle
12•lftherios•46m ago•2 comments

Building the Data Backbone Behind Real-Time Financial Systems

https://alltechmagazine.com/real-time-financial-data-architecture-gayathri-balakumar/
1•dekhna•53m ago•0 comments

I built the LocalStack alternative I wanted (real containers, not mocks)

https://floci.io
2•hectorventura•53m ago•0 comments

New Google Networks Tuned Up for GenAI Inference and Training

https://www.nextplatform.com/connect/2026/04/28/new-google-networks-tuned-up-for-genai-inference-...
1•rbanffy•53m ago•0 comments

Software Heritage

https://www.softwareheritage.org/
2•p4bl0•55m ago•0 comments

Start with the sensors, then design the rest: How Zoox built its robotaxi

https://arstechnica.com/cars/2026/04/start-with-the-sensors-then-design-the-rest-how-zoox-built-i...
1•rbanffy•56m ago•0 comments