Show HN: The current sky at your approximate location, as a CSS gradient (sky.dlazaro.ca)

418 points by dlazaro 7h ago 85 comments

Show HN: I built a platform to connect with future peers before you start (findeaze.com)

2 points by suhaibkhan699 3h ago 0 comments

Show HN: Runtime – skills-based browser automation that uses fewer tokens (github.com)

2 points by tito777 3h ago 0 comments

Show HN: Kitten TTS – 25MB CPU-Only, Open-Source TTS Model (github.com)

967 points by divamgupta 3d ago 355 comments

Show HN: Trayce – Burp Suite for developers (trayce.dev)

77 points by ev_dev3 2d ago 14 comments

Show HN: Browser AI agent platform designed for reliability (github.com)

72 points by ogandreakiro 2d ago 32 comments

Show HN: Synchrotron, a real-time DSP engine in pure Python (synchrotron.thatother.dev)

79 points by andromedaM31 1d ago 4 comments

Show HN: I made a Ruby on Rails-like framework in PHP (Still in progress) (github.com)

3 points by al3rez 8h ago 2 comments

Show HN: Stasher – Burn-after-read secrets from the CLI, no server, no trust (github.com)

70 points by stasher-dev 2d ago 53 comments

Show HN: Octofriend, a cute coding agent that can swap between GPT-5 and Claude (github.com)

98 points by reissbaker 2d ago 30 comments

Show HN: Aha Domain Search (ahadomainsearch.com)

45 points by slig 5d ago 29 comments

Show HN: Sinkzone DNS – Forwarder that blocks everything except your allowlist (github.com)

95 points by dominis 3d ago 43 comments

Show HN: Event Sourcing and AI (eventsourcing.ai)

3 points by goloroden 13h ago 1 comments

Show HN: I spent 6 years building a ridiculous wooden pixel display (benholmen.com)

1125 points by benholmen 5d ago 142 comments

Show HN: AI feedback on system design diagrams (system-design-6m8.pages.dev)

3 points by gpe 13h ago 5 comments

Show HN: Keywords for Self-Talk (plastithink.com)

3 points by andsko 13h ago 3 comments

Show HN: I've been building an ERP for manufacturing for the last 3 years (github.com)

320 points by barbinbrad 4d ago 157 comments

Show HN: Aura – Like robots.txt, but for AI actions (github.com)

40 points by OsmanDKitay 3d ago 33 comments

Show HN: Whittle – A shrinking word game (playwhittle.com)

138 points by babel16 4d ago 56 comments

Show HN: Tiered storage and fast SQL for InfluxDB 1.x/2.x (historian.exydata.com)

4 points by ignaciovdk 22h ago 0 comments

Show HN: An open-source e-book reader for conversational reading with an LLM (github.com)

85 points by takigon 3d ago 61 comments

Show HN: I made a safe anonymous message app (subrosa.vercel.app)

3 points by martianmanhunt 19h ago 4 comments

Show HN: GPT-5 Document Retrieval – AI Assistant with Inline Citations (smartresearch-ai.com)

5 points by ben011 1d ago 0 comments

Show HN: LLM from URL –– A free AI chat completion service directly from URL (818233.xyz)

7 points by yvonuk 1d ago 4 comments

Show HN: When is the next Caltrain? (minimal webapp) (erikschluntz.com)

42 points by eschluntz 3d ago 27 comments

Show HN: I made FiscalBud to send invoices fast and worldwide in 77 languages (apps.iridesk.com)

3 points by drex91on 23h ago 0 comments

Show HN: Sidequest.js – Background jobs for Node.js using your database (docs.sidequestjs.com)

88 points by merencia 5d ago 28 comments

Show HN: Tiny logic and number games I built for my kids (quizmathgenius.com)

87 points by min2bro 5d ago 37 comments

Show HN: HMPL – Small Template Language for Rendering UI from Server to Client (github.com)

16 points by aanthonymax 3d ago 7 comments

Show HN: Selfhostllm.org – Plan GPU capacity for self-hosting LLMs (selfhostllm.org)

7 points by erans 1d ago 3 comments

Show HN: BaaS to build agents as data, not code (github.com)

5 points by ishita159 1d ago 0 comments

Show HN: Aegis – A framework for AI-governed software development (github.com)

5 points by chavezabelino 1d ago 0 comments

Show HN: Spellbook, a system package manager written in Elixir (spell-book.run)

3 points by tcmart14 1d ago 0 comments

Show HN: Bringing Tech News from HN to My Community (sh4jid.me)

3 points by sh4jid 1d ago 2 comments

Show HN: Kimu – Open-Source Video Editor (trykimu.com)

103 points by robinroy03 5d ago 35 comments

Show HN: I built a (legit) AI mortgage document analyzer that saves you money (oxford.loan-estimate-analysis.morfi.com)

3 points by matthew-morfi 1d ago 0 comments

Show HN: I built a tool that lets you summon AI in any app or website (useinset.com)

3 points by blaumaus 1d ago 2 comments

Show HN: I built a free fast-paced match-3 game I think you'll like (sqrz.app)

4 points by jasonjmcghee 1d ago 1 comments

Show HN: Starter Repo for Rust CLI with built in database (deebkit.com)

2 points by nickisyourfan 1d ago 0 comments

Show HN: Wordle" with the first 1000 Digits of Pi (subjectle.com)

4 points by bezo226 1d ago 2 comments

Show HN: FFlags – Feature flags as code, served from the edge (fflags.com)

41 points by tushr 5d ago 36 comments

Show HN: I completed mobile fax app in 60 hours

3 points by tntpreneur 1d ago 6 comments

Show HN: Stagewise (YC S25) – Front end coding agent for existing codebases (github.com)

46 points by juliangoetze 4d ago 50 comments

Show HN: Fiig – Reimagining Google scholar for AI research and PDF annotation (app.ubik.studio)

3 points by ieuanking 1d ago 1 comments

Show HN: A light GPT-5 vs. Claude Code comparison (charlielabs.ai)

13 points by neom 2d ago 0 comments

Show HN: I built a tool to replace capcut audio transcription (meetcosmos.com)

26 points by correa_brian 4d ago 10 comments

Show HN: I built a simple tool to automate data into Google Sheets and BigQury (syncrange.com)

3 points by RyanDavid 1d ago 0 comments

Show HN: Chilli – A lightweight microframework for CLIs in Zig

7 points by habedi0 3d ago 0 comments

Show HN: Rust framework for advanced file recognition and identification (crates.io)

44 points by reimisdev 2d ago 7 comments

Show HN: I built a text-based birthday reminder app (birthdays.app)

46 points by samfeldman 4d ago 31 comments

GPT-5 System Prompt

28 georgehill 9 8/9/2025, 6:48:24 AM github.com ↗

Comments (9)

dgreensp · 4h ago

> Never place rich UI elements within a table, list, or other markdown element.

> Place rich UI elements within tables, lists, or other markdown elements when appropriate.

crazygringo · 4h ago

How does a prompt this long affect resource usage?

Does inference need to process this whole thing from scratch at the start of every chat?

Or is there some way to cache the state of the LLM after processing this prompt, before the first user token is received, and every request starts from this cached state?

mdaniel · 4h ago

My understanding is that is what the KV cache does in models serving. I would imagine they'd want to prime any such KV cache with common tokens but retain a per-session cache to avoid leaks. It seems HF agrees with the concept, at least https://huggingface.co/docs/transformers/kv_cache#prefill-a-...

mdaniel · 4h ago

It's a good thing people were enamored of how inexpensive GPT-5 is, given that the system prompt is (allegedly) 54kb. I don't know how many tokens that is offhand, but what a lot of them to burn just on setup of the thing

btdmaster · 3h ago

I might be wrong, but can't you checkpoint the post-system prompt model and restore from there, trading memory for compute? Or is that too much extra state?

mdaniel · 3h ago

My mental model is that the system prompt isn't one thing, and that seems even more apparent with line 6 telling the model what today's date is. I have no insider information but system prompts could undergo A/B testing just like any change, to find the optimal one for some population of users

Which is to say you wouldn't want to bake such a thing too deeply into a multi-terabyte bunch of floating points because it makes operating things harder

Tadpole9181 · 2h ago

54,000 bytes, one byte per character. 4 characters per token (more or less). Around 13,000 tokens.

These are NOT included in the model context size for pricing.

TZubiri · 5h ago

These are always so embarassing

NewsaHackO · 4h ago

It's because they always put things that seem way to specific to certain issues, like riddles and arithmetic. Also, I am not a WS, but the mention of "proud boys" are things that can be used as fodder for LLM bias. I wonder why they even have to use a system prompt; why can't that have a separate fine-tuned model for ChatGPT specifically so that they don't need a system prompt?