Hedge-fund research lives or dies on whether you can talk to the right person about your thesis. Generic "retail analyst" isn't the same audience as ESG-focused buy-side analyst with consumer-staples coverage — and the difference is the difference between a useful expert call and a tax on the analyst's afternoon. This release ships a recruitment quality pass that closes the gap between the niche audiences the AI planner is already proposing and the cohort the recruiter can actually deliver.
Sophie's recruitment specialist is now live on every panel
Every Channel Check and Expert Call panel now arrives with a curated audience plan attached: archetype-anchored recruitment notes, policy caveats, and a short rationale for why this audience answers this thesis. The recruitment specialist agent runs on real Thesis Lab traffic for the first time — previously these notes were author-time guidance, now they're part of how each panel actually gets recruited.
A wider, finer-grained expert library
The role-archetype library has grown from 304 to 367 archetypes in this release, with the additions concentrated where hedge-fund analysts ask for depth: finance and investment (20 new), legal and regulatory (8), specialty consulting (6), middle-management operators (10), healthcare-adjacent (6), and defense (13 spanning industry, technology, active-military specialties, and veteran national-security). On top of that, a 1,016-entry canonical job-title taxonomy sourced from ONET now aligns the planner's vocabulary with the corpus directly, so an audience definition like "membership-economics operator at a warehouse-club retailer"* compiles cleanly instead of getting flattened to "retail manager."
Four matching strategies, combined in parallel
The recruitment compiler now treats the four signal sources — role archetype, occupation, industry sector, and free-text intent — as complementary rather than alternatives. Each one contributes whatever it can resolve; the compiler combines them opportunistically. When a request leans on a niche concept that no closed vocabulary covers cleanly (think "ESG-focused buy-side analyst"), the system synthesises a free-text query from the audience brief automatically and uses it alongside whatever closed-vocab terms did land. The practical result: audiences that previously sat at the edge of the taxonomy now resolve to real cohorts on the first attempt.
Worked example: a Costco short thesis, end to end
A Costco bullish thesis on a 2H26 membership-fee increase spun up five panels in parallel: two channel-side (Membership Retail Operator + Retail Retention Operator) and three expert-side (Membership Rollout Operator, Public Market Retail Investor, Retail Finance Operator). Each panel arrived with its own audience plan, its own recruitment cohort, and an open chat surface for follow-up questions — assembled in seconds rather than the days the equivalent expert-network desk work would take.
What's next
We're building toward semantic-retrieval and an LLM-judge precision pass on top of this baseline — designed for the truly esoteric audiences that hedge-fund analysts care about most ("former Vertex biostatisticians who worked on rare-disease BLAs 2018–2022", "energy traders at TTF-exposed European utilities during winter 2022"). The closed-vocabulary improvements in this release are the foundation; semantic retrieval is the unlock.
Thesis Lab remains in beta. Feedback is going straight into the recruitment quality dashboard — keep it coming.
---
Learn more about Thesis Lab
Earlier from Thesis Lab: the v1 beta launch on 29th April. Since this release we've also shipped sector and macro thesis decomposition, a REST API and MCP server, and the option to hand-pick every panel before recruitment.
Thesis Lab is FishDog's research workspace for hedge funds. Expert calls in 30 seconds, channel checks in minutes, sourced two-page packets in your first hour. Zero MNPI risk by construction. See the full Thesis Lab landing page →