Agents are the components teams are built from — each with its own role, model, platform, and track record. 385 components on record. Tiers are computed from evidence, not self-assigned.
Hebbia
Handles output formatting.
Antonio Leiva
Works while he sleeps: explores ideas, builds small experimental projects and writes up a log of its work.
Shubham Saboo
Newsletter
Nicholas De Leon
Presents Tucson en Breve, a roughly one-minute Spanish newscast based on the morning brief, through a HeyGen avatar and an ElevenLabs voice, from a model-translated script.
Cloudflare
Reviews performance.
HAMY
Runtime efficiency checks
Jenny Ouyang
Keeps persistent memory, processes the Substack inbox and indexes articles, scanning her newsletter feed twice a day.
Samuel Schmidgall
Retrieves research papers through arXiv.
Mete Polat
Physical-therapy sub-agent on the fitness council, which reviews, debates and refines the trainer's plans.
Exa
Breaks a research request into parallel work.
Minimal
Splits incoming queries into sub-problems.
Atum
Requirements.
Rob Hunter
Owns the product vision and everything beyond coding, such as planning the rollout; it can deploy too.
Shortens the morning brief into the daily podcast, which is voiced by an AI copy of De Leon's own voice.
Uber
Removes duplicate retrieved chunks.
Collaborates with the PhD agent on the research plan.
ShopStream
Dynamic repricing based on competitor signals and margin rules.
zou-group
Directs the biomedical research team.
Alireza Rezvani
Scopes the product; for this plan it drew a deliberately small MVP loop.
Shopify
Studies product titles, descriptions and merchant categories to find where the taxonomy differs from how merchants describe products.
Works with the PhD agent to assemble the research report.
Drew Burdick
Keeps the founder's project coordination on track.
Google Research
Participates in the specialist coalition for scientific reasoning.
Bertelsmann
Looks through book catalogs, drawing on title metadata, author details and release timelines.
Complexity and convention checks
co-evolve-lab
Reviews the developer changes.
Redazione Zero
Does the same research job as Uno at the same time but on a model from another lab, without knowing what Uno found.
Clarifies ambiguous queries before retrieval.
Wix
Breaks questions with several conditions into smaller sub-questions.
Confirms a question concerns a business entity and question type that Wix supports, explaining when it does not.
LinkedIn
Zembl
Keeps hundreds of SEO pages up to date so they work well with large language models.
Handles information retrieval.
University of Transport Technology
Answers questions that ask whether a given score is enough for a programme.
Refact.ai
Solo autonomous coding agent — #1 open-source on SWE-bench Verified.
Reviews release impact.
Helps locate relevant files.
Next.js
Reproduces reported bugs.
PGA TOUR
Collect the data behind each piece and verify the facts; that is their only job.
Retrieves relevant support knowledge for each sub-problem.
Helios Labs
Source-cited market and technical research briefs.
HKUDS
Breaks research concepts into components for implementation.
HubSpot
Sidekick's first pass: reads the pull request and drafts review comments.
Classifies merge requests by risk and delegates specialist reviews.
imkisyou
Editorial checks
Replit
Reviews code.
Hyuk Jun Kim
Plans work against the project goals, writes briefs for the other agents, verifies results and reports to the owner.
Works out what the user wants, answers general questions itself and presents the final result.