OpenAI's latest model release with prompt-length and token-spend debate.
Show EvidenceHide Evidence
Replies compared GPT-5.6 with Codex and Claude Code, and debated whether shorter prompts really save tokens.
Issue / 2026-07-09
A daily board of tools, apps, and references that Hacker News readers pulled into view on 2026-07-09. Each row keeps the original HN thread close to the claim.
What surfaced that day
OpenAI's latest model release with prompt-length and token-spend debate.
Replies compared GPT-5.6 with Codex and Claude Code, and debated whether shorter prompts really save tokens.
OpenAI's work-focused ChatGPT/Codex UI consolidation.
Commenters focused on the confusing split between ChatGPT Work and ChatGPT Codex, calling the app change hard to follow.
Meta's model API release pitched at lower-cost inference.
Discussion compared Muse Spark 1.1 with Grok 4.5 and highlighted the cached-input price/performance tradeoff.
An AI text detector and browser extension for written content.
One commenter said 'Pangram does work' and described it as useful for spotting AI-written text.
A lightweight local runner for GLM 5.2 on slow computers.
A commenter said the readme benchmarks looked great across different hardware and reported good times on the Colibri runs.
An API that turns websites into structured data and screenshots.
A user said Context.dev was a great product that made a hard problem disappear, then recommended Intercept as an alternative.
A typed proxy API that maps the real transport layer of websites.
One commenter said "Have a look at Intercept" and described it as a typed TypeScript proxy API that reverse-engineers website traffic.
An ACME DNS helper for issuing certs to internal services.
A commenter said, "I do this exactly, using ACME DNS," in the internal certificate discussion.
An agent-first CLI for scientific literature and evidence.
Commenters called the atlas neat, asked for search-term filtering, and described it as useful for exploring a corpus.
An interactive analog watch game that trains clock reading.
Replies called it fun and said the practice rounds were needed to stay sharp.