COMMUNITY PICK · 3 NOVEMBER 2026
Tinyfish's founder: five lessons from building for agents
@Saboo_Shubham_ shared five takeaways from talking to the founder of Tinyfish: agents are becoming major users of the web, so sites will turn agent-friendly; human approval steps are a phase, not the end state; subscriptions don't suit agents, so Tinyfish charges per task from a prepaid wallet; logins sit in a vault the model never sees, with a separate model checking each action; and the founder runs 12 Claude Code sessions in parallel. These are a vendor's opinions.
Why it mattersTwo ideas to copy now: keep passwords out of your agent's reach (a vault it can use but not read), and add a second check on any action that spends money or sends something.
Curated by Sara Simeone as part of our human editorial coverage. The original sources are linked below; curation is distinct from a Hype or Fact? assessment.
Sources
Related AI tools
How we knowCompany figures are labelled as their own claims; anything we couldn't verify is labelled as a report or a claim.
MORE FROM 3 NOVEMBER 2026
- Anthropic's IPO filing: it may 'imminently' hit its own threshold for AI-driven progress
- 'Weight-space programming': GPT-6 Astra edited a small model's weights and lifted its maths score from 7.9% to 48.1%
- GDPx: the best AI agent finishes only 12% of 206 hard, high-value work tasks
- OpenAI's always-on agent, dots, is moving to GPT-6.1 Terra to reach more people
- Altman: the model behind OpenAI's Navier–Stokes result is coming 'very soon' (no date)
- Hydrogen-1.0: an open-weight science model that its maker says beats GPT-6 Astra on research questions
