chat windows, sparkles, and agent swearing
gm legends
Today: an app builder that gave up on having its own AI, a browser extension that won't let you get away with just watching the tutorial, and analytics that count how often your team swears at Claude.
Also: Vercel Day is tomorrow, so last call to get on the board.
One day. Unlimited launches. Zero brakes.
Hypership Day is Tuesday, September 30. Launch in the morning, read what people tell you, build the thing they asked for, and launch again. As many times as you can stand, until midnight.
It's open to anyone who hasn't launched in the last six months and thinks they can ship faster than everyone else. We've partnered with TypeSafe AI, makers of Jev, a frontier model that answers at real-time speed, and Supabase which felt like the right companies for a day about velocity. Prizes include early access to Jev, credits, swag, and a few we're keeping quiet about.
Not launching? Turn up anyway. The whole thing only works if people are trying the products and saying what's missing, and on the day your feedback is somebody's next feature.
Floot MCP — Build and ship web and mobile apps inside Claude or ChatGPT
Floot MCP hands Claude or ChatGPT a cloud workspace that already has a database, user logins, hosting and a live URL wired up, so the app gets built, published and updated without anyone leaving the chat. Floot spent its first year building an app builder with its own agent in it, and Yuj's read now is that Claude and ChatGPT got good enough at writing software that a second agent stopped being worth having. Your own subscription pays for the thinking, so they charge a flat platform fee with no metering on top, which is a lot of margin for an AI company to hand back, and the same project ships to web, iOS and Android.
Arcjet — Your agents stopped answering and started doing
(sponsored)Your agents call tools, write to databases and move money, so a wrong answer is now a wrong action, which is a considerably worse meeting. Arcjet runs policy before every LLM call, tool call, query and API request: prompt injection, sensitive data, bots, rate limits, spend quotas, and your own rules in Rego on top. It reads the OpenTelemetry you already emit, so switching it on is one environment variable, no code changes and nothing to deploy. Each check also reads the earlier steps in the same workflow, so an agent's tenth action gets judged against the first nine and not on its own. JavaScript, Python and Go SDKs all at 1.0, agent framework guards included, free tier to start on.
Scholé Learn by Building — Learning should be more than "click next"
Scholé is a browser extension that sits beside you while you use a real tool and makes you do the work. It watches the screen, ticks off the task list, and drops little sparkles where you're meant to look next when you're stuck. Vini and the team are AI-for-education researchers who use AI constantly and have landed on an uncomfortable point: you can read about building an agent, watch someone do it, get ChatGPT to explain it beautifully, and none of that is learning it, but if the AI builds the whole thing for you then you haven't learned it either. They track where you got stuck and what you misunderstood, then try a different explanation until it lands.
Opaline — PostHog for team Claude Code and Codex sessions.
Opaline is message-level analytics for your team's Claude Code and Codex sessions, so token cost, time and skill usage per message. Evren's argument is that the existing tools are single-user and shallow, fine for one developer and useless once a team is running thousands of sessions a week. The honest detail in the launch is that it catches every "you're absolutely right" from Claude and every "I know, you f*cking idiot" typed back by somebody out of patience, and Evren's defence of that is that the more you swear, the more data you have to steer your agents with. His own team's default model changed twice in a month, and that shift is the other thing this measures.
Daily Top Products
