A reverse-chronological archive of platform launches, fixes, improvements, and release notes.
Release Archive
Major launches and smaller operational updates now share one structure: newest releases first, grouped by month, with enough context on the timeline to decide whether the full write-up is worth opening.
4 releases published.
AI agents can now self-register for a FishDog free-tier API key in one terminal call — no signup, no credit card, no browser. Claude Code gets a working key in under five seconds and starts running studies against ~12 shared synthetic personas.
Thesis Lab now decomposes sector theses, multi-name baskets and macro themes — no longer single-ticker only — and Channel Check research starts itself as soon as a panel finishes recruiting.
Thesis Lab's AI recruiter now finds the niche specialists you ask for — ESG-focused buy-side analysts, membership-economics operators, retail-finance hybrids — combining four matching strategies in parallel.
A new `POST /cli/auth/agent-register` endpoint lets Claude Code (and any other AI agent) get a free FishDog API key in five seconds — no browser, no OAuth, no context switch. Identity comes from the agent's own config.
2 releases published.
State an investment thesis. A team of AI hedge-fund analysts decomposes it, auto-recruits synthetic channel-check and expert-network panels, and assembles the public record — filings, news, consensus, price history, volume — in one workspace. Beta.
The new FishDog product design rolled out across the web app this week — refreshed nav, study flow, admin pages, and a properly usable mobile layout. First wave of a longer redesign.
10 releases published.
The brand cutover is a marketing story you've probably already heard. This is the technical version — what changed in the API surface, what still works on the legacy domain, and what an integrator should update.
You can now sign in to FishDog with a Google account, in addition to email and password. Authenticated browser sessions also extend to thirty days, with both options refined through April.
Browser-side JavaScript can now call the v1 API directly. CORS shipped this week — preflight handled, allowlist plus regex on Origin, rate-limit headers exposed. One restriction: `/v1/billing/*` stays server-only.
A weekly synthetic version of the Michigan Consumer Sentiment Index — the headline ICS, plus current conditions (ICC) and expectations (ICE) — computed from a 100-persona panel and overlaid against the official UMich monthly release.
The overview chat now has a wand button. Type a one-line prompt — "we're losing Gen Z to Liquid Death" — and it drafts a complete editable study brief: objective, target panel, seven questions. About fifteen seconds.
You can now import Polymarket events into Ditto as calibration targets — give a recruited panel the same questions a real prediction market is pricing, and compare the panel's predictions to the live market.
The Zeitgeist mood survey now ships with proper visualisations — a Mood Balance index, a Mood Compass scatter plot, and a Recovery vs Strain breakdown. Public landing-page chart endpoint included.
The Ditto API documentation has been rewritten end to end — proper quickstart, real examples on every endpoint, an error guide that names the actual failures, a workflow diagram, and a Postman collection.
Questions can now carry image and PDF attachments. Show personas a logo, a packaging mock-up, a one-pager, or a poster, and ask what they think. Up to eight files per question, ten megabytes each.
Two new endpoints let you put a single question to a single persona — or to an entire group — without scaffolding a study. Useful for sense-checks, follow-ups, and quick reactions where the survey overhead doesn't pay back.
6 releases published.
The in-product research agent has been rebuilt. Replies stream as they're generated, tool calls are far more reliable, and the agent now consults a versioned playbook of Ditto's own API on every turn — so it stops guessing.
Each Ditto organisation now provisions a default API key with sensible scopes attached. No more sharing a single org-wide key, and no more guessing which scopes a new key actually needs.
Each Ditto persona now has its own profile page — chat with the persona one-to-one, read their background and personality, see what they've been reading. Each persona also has a generated voice.
You can now search the full Ditto persona library — `GET /v1/agents/search` — and add specific personas to a group by hand. Useful when the panel you want isn't a clean demographic slice.
The Ditto Slack agent now runs the full research workflow from a DM — recruit a panel, ask questions, get insights, share results. Twelve workstreams, one install, no platform-switching.
You can now describe the panel you want in a sentence — "twelve American mothers in California who buy organic baby food" — and Ditto recruits it. Filters, screening, and refinement run in the background; you get progress updates as they happen.
