
ChatGPT Sites on Hacker News: What Builders Should Know
Hacker News hit ~209 points on Sites in ChatGPT. What OpenAI docs say about custom domains, Sign in, storage, and when .chatgpt.site is the wrong bet.

Expert profile
Founder & AI Product Leader
Yash Thakker is a Generative AI expert with over 12 years of experience in product leadership and technical strategy. As the founder of explainx.ai, he has taught over 300,000 learners and built AI platforms serving millions of users globally. He specializes in Agentic AI, Multimodal RAG, and the intersection of LLMs with consumer hardware.

Hacker News hit ~209 points on Sites in ChatGPT. What OpenAI docs say about custom domains, Sign in, storage, and when .chatgpt.site is the wrong bet.

Install and build Claude Code mods in the desktop app: Plugins menu, scopes, local folders via CLAUDE_CODE_PLUGIN_DIRS, a tested SVG limits ring.

Anthropic launched Claude Frontier Academy on Oct 2, 2026: $100M, 12-week residency, nomination-only. What it is, who gets in, and alternatives.

Anthropic highlighted Opus 5.5 builds: a watermill, embossed cards, a 3D Milky Way to Jupiter site, browser water. Nine community clips and a build guide.

Opus 5.5 ranks first of 253 models on the Epoch Capabilities Index at 167.35. What the score includes, where GPT-6 Astra leads, and what to pick.

Cloudflare Sandbox SDK 1.0: Durable Object owns ctx.container, 6x faster starts, filesystem snapshots, Containers pricing, Artifacts fit.

Cloudflare Web Search API runs through AI Gateway with Exa, Linkup and Ceramic. Code samples, billing at partner list prices, logging, and limits.

Cua Spaces launched Oct 2, 2026: free, source-available desktops for agents across your computers. Teleport, Keyvault, pricing, and how it differs.

Can DwarfStar ds4 run DeepSeek V4 Flash on your Mac this week? Quant, tok/s, and Claude Code setup, checked against the README.

zkAPI lets you pay AI services from a private balance so providers cannot link requests to you. Mechanics, experimental status, privacy limits, builder uses.

Google Research's Cogentic harness used Gemini on five open TCS math problems. Multi-agent prove–verify, expert review, not Meta Muse Spark.

The viral 81% Astra cost cut is wrong. Artificial Analysis puts GPT-6.1 Sol at $0.72 vs Astra $3.26 per Index task — ~78% cheaper, not list-price math.