01

cat architecture.md

MINT runs a local-first research infrastructure on a Mac Studio. FeedMe reads hundreds of sources three times a day and ranks what matters to the lab. Yesterday in AI is written from its shortlist, with Seth choosing and approving every issue. A searchable paper corpus, with its own paper-fetching and review services, sits behind the lab's Slack agents and a read-only server for collaborators' assistants. Shared services handle credentials, model access, backups and the public route in. Public pages are static files deployed through GitHub Pages.

56
Loaded launchd jobs
9,523
Corpus documents
620,584
Searchable chunks
49
Lab repositories
Live documentation. Values marked by the status dot come from a public, scrubbed snapshot of the running system. Minty's canonical configuration and local services generate the snapshot twice daily. Last refresh: November 15, 2026.
02

trace content-pipeline

FeedMe is the shared collection and ranking layer. The full 04:00 run starts all seven fetchers at once, drops repeats and stale items, scores every new item against the lab's interest profile, summarizes long pieces, and groups the survivors into stories on the day's slate. The 12:00 and 20:00 runs are top-ups with six fetchers that merge new stories into the same slate. Items count as seen only once the slate accepts a run, so a failed run is fetched again. A watchdog checks the morning run thirteen times before 08:00 and restarts it if it stalls.

01 Collect Seven fetchers
  • X
  • 5 Bluesky feeds
  • ~380 feeds, journals, podcasts
  • Newsletter inbox
  • arXiv + PhilPapers
02 Process FeedMe
  • 04:00 full run
  • 12:00 top-up
  • 20:00 top-up
Drop repeats, score, summarize, group into stories. A watchdog checks the morning run until 08:00.
03 Route Two outputs
Day's slate
up to 300 sources
Morning picks
20 to 25 to #firehose
Yesterday in AI daily, from the slate
  1. SethMarks the slatein the curator app
  2. Selector19:30 to 23:30, about 40 sources, continuity check
  3. SethApproves the pitchproposal at 23:00
  4. Writer + deskdraft, Read mores, edits, preview
  5. Releasewebsite, one email per subscriber, #mint-community
Slack from the morning picks
  1. Picks + sharesmorning picks, and the share bot every 30 s
  2. #firehosethe lab's shared stream
  3. Paper botevery 5 min, fetches papers posted there
  4. #paperspapers cross-posted; requests filed for any it can't get
Corpus from #papers, a local folder and a Drive inbox
  1. Intake#papers every 10 min, folder overnight, Drive every 30 min
  2. Fetchuniversity library and a local model
  3. Analyze41 questions and six summaries per paper
  4. Single writerone locked writer; archive to Drive and Zotero
  5. Serve@Minty, the corpus server, and the Corpus Map (Sundays)
How an item travels: collected, ranked by FeedMe, then routed to the newsletter, Slack and the corpus. Amber steps wait for Seth.

Source adapters

X timelines
BirdClaw reads the followed and recommended timelines through a signed-in session, with fallbacks when a live read fails. It expands busy conversations into threads, stopping at the first refusal or rate limit, and fetches the full text of articles that short posts point to.
Bluesky feeds
An authenticated AT Protocol client reads five feeds and expands threads before writing normalized items.
Feeds, journals and podcasts
About 380 web feeds, journals and podcasts are read each run, keeping recent unseen entries. Journals get longer windows and a cap on new articles, and the morning run adds podcast captions. A weekly source finder reads roundup posts and adds high-confidence recurring sources.
Newsletter inbox
The Gmail API reads newsletter mail from a dedicated inbox, strips tracking, and filters out account and promotional mail. Full text for cut-short or paid articles comes through web requests or signed-in browsers, including the university library.
arXiv
New paper listings are read every run, dropping papers already seen. Each paper gets its authors and institutions, and paper posts by trusted researchers are marked to pass scoring.
PhilPapers
The morning run reads recent PhilPapers additions alongside arXiv, across the lab's philosophy and technology-ethics categories.

Products and agents

Yesterday in AI
Seth marks stories on the day's slate in the curator app. An evening selector chooses about forty sources and checks each against earlier issues; a 23:00 plan goes to Seth's pitch room, and nothing is written until Seth approves it. Agents then write the digest, research and edit the Read mores, and send Seth a preview. Only an approved issue is published on the website, emailed once to each subscriber, and linked in #mint-community.
Slack picks
After the morning run, two model passes narrow the passed items to a varied 20 to 25, which are posted to #firehose with full-text summaries. Links shared by Seth and lab members are scored and posted there too.
Corpus Agent
news-pdf-bot fetches papers linked in #firehose and cross-posts them to #papers, filing a request when it cannot get a file. The corpus pipeline ingests them, and the persistent corpus-agent answers research questions from the corpus through @Minty in Slack.

A sign-in keeper holds the browsers that read paid sites, the university library and X, tests each login at 09:10 and 17:10, and emails Seth when one needs renewing. Each source is judged against its own recent normal, and Seth is told once when an important source fails and once when it recovers.

03

trace corpus-ingestion

Paper ingestion is a queue-backed set of focused services. Papers arrive from the #papers channel, a local intake folder and a shared Google Drive inbox; papers that an ordinary download cannot get are found by a separate fetching service working through the university library. Each paper is fingerprinted so repeats are set aside before any costly work, then identified, converted, embedded and analysed. One writer owns every database change, and the archive keeps the original PDF on Drive and an entry in the lab's Zotero library.

  1. corpus-intake-localWork the local intake folder overnight, every 5 minutes.
  2. corpus-intake-slackTake papers from Slack from 08:00 to midnight, every 10 minutes.
  3. corpus-ingest-handoffHand unfinished work over at midnight and top up the overnight queue.
  4. corpus-analyzeAnswer 41 research questions and write six summaries per paper.
  5. corpus-db-writerCheck each handoff against its archived files, write it, and read it back.
  6. corpus-tidyClear intake copies once the archived copy is confirmed.
A single database writer is deliberate: every change takes the same write lock, and a handoff whose files or passages do not match is refused and kept for a person to settle, so searches only ever see papers that match their archived files.
04

du -sh corpus/

9,523
Documents
8,761
PDF-backed records
762
Native web sources
620,584
Chunks
8,950
Analyses
8,950
RAPTOR summaries
16
Topic clusters

The corpus contains both PDF-backed papers and sources captured directly from the web as Markdown. Each document is represented at several levels: bibliographic metadata, full text, searchable chunks, embeddings, structured research-question analysis, and hierarchical summaries. Every Sunday the corpus is regrouped into topic clusters, and each paper keeps a list of its ten closest neighbours. The current document embedding policy is Voyage voyage-context-3, 2048d.

Explore the public projection in the Corpus Map. The public map exposes research structure and links, while the working corpus remains on lab infrastructure.

06

cat AGENTS.md

Minty is the lab's research-management agent. The root AGENTS.md is the always-on contract for identity, operating boundaries, verification, and collaboration. Each persistent workstream keeps one THREAD.md with its rules, sources, decisions and blockers, so a new agent can pick it up.

Canonical extensions
.codex/ owns skills, commands, workers, tasks, and hooks.
Compatibility mirrors
.agents/ and .claude/ are generated from canonical files.
Agent identity
Runtime agents use Lab-Minty-[hex].
21
Skills
22
Commands
6
Worker specs

The old root fact digest and session database are retired archives. Durable lessons belong in the active runtime's memory, and exact history stays in that runtime's native transcript store.

07

launchctl list | sort

launchd owns persistent and scheduled services. The table below is generated from Minty's canonical daemon registry and reconciled with the jobs currently loaded on the host. 54 user jobs and 2 system jobs are loaded.

Service Purpose Schedule Observed state
loading snapshot Reading the public infrastructure inventory. - • Loading
08

calendar daily-jobs

Time (ET) Service Work
04:00 feedme-daily Full source collection, scoring and the day's slate, then the Slack picks
04:42 to 08:00 feedme-daily-watchdog Thirteen checks on the morning run; restart it if it stalls
05:15 minty-chatbot-sync Rebuild the site assistant's records when they change
09:10 and 17:10 yinai-health-check Test the paid-site, library and X sign-ins
12:00 and 20:00 feedme-incremental Top-up runs that add new stories to the day's slate
19:30 to 23:30 yinai-selector Choose about forty sources for the next issue
23:00 yesterday-in-ai Plan the issue for Seth's approval
After approval yesterday-in-ai-write-consumer
yinai-proposal-release
Write, edit and preview the issue; publish and send once approved
00:00, 06:00, 12:00, 18:00 rclone-backup Back up lab data; the 06:00 run archives the corpus and mirrors off-site
06:00 and 18:00 guide-updater Refresh and publish this page's system snapshot
Sunday 04:00 paper-map-updater Regroup the corpus if due and redraw the Corpus Map
Sunday 05:00 slack-index Refresh the Slack search index
Continuous corpus-* Queue-backed paper intake, analysis, and writes
Persistent corpus-agent Answer grounded corpus questions and add tagged papers in Slack
Every 30 seconds share-to-minty Score and post links shared in Slack
Every 5 minutes news-pdf-bot Fetch papers linked in #firehose

Exact launch intervals and loaded state are listed above. The schedule table highlights the main lab-facing rhythm rather than duplicating every watchdog and maintenance interval.

09

help lab-members

Bots

@Minty
Ask research questions answered from the corpus, choose a mode and model, or tag it on a paper to add it to the corpus.
@minty-search
Search the lab's Slack history by meaning. It answers citing message links and says when the evidence is thin.
Share bot
Drop a link in a share channel and it reads, scores and posts it to #firehose, where the newsletter and the corpus pick it up.
Paper bot
Post a paper link in #firehose and it saves the PDF to the thread and the corpus, or says why it could not.

Slack channels

Channel Role
#mint-community Lab discussion, shared links, and each Yesterday in AI issue link.
#firehose Morning news picks and shared links; papers posted here are fetched.
#papers Paper intake for the corpus, with ingestion replies.
10

inspect agent-runtime

Scheduled agents send model calls through a shared router that rotates through the lab's model sign-ins and rests any that hit a usage limit; a probe checks every sign-in every ten minutes without spending a call. Jobs get credentials by name from a signed helper that keeps them out of code and logs. Deterministic services use ordinary code where a model adds no value; research synthesis and judgment-heavy work use higher reasoning effort.

Task class Reasoning effort Typical work
Intelligence high Research, synthesis, review, and complex judgment
Functional low Implementation and bounded operational tasks
Deterministic low Mechanical checks where a model is still required
Verification is part of the runtime contract: agents must collect fresh evidence before reporting state, preserve concurrent work, and keep credentials out of public artifacts. Agents prepare; Seth approves anything that is published or sent.
11

ls integrations/

Slack
Research questions, link and paper intake, search, news picks, and newsletter links.
Listmonk and PostgreSQL
Newsletter subscriptions and email delivery records.
Google Drive and Zotero
The paper archive, a shared intake inbox, off-site backups, and the lab's reference library.
GitHub
49 repositories in the mint-philosophy organization.
GitHub Pages
Static deployment for mintresearch.org and its public artifacts.
Academic APIs
arXiv, PhilPapers, Crossref, Semantic Scholar, and open-access lookups, plus the university library for paywalled papers.
Model providers
Language models for scoring, writing and research, reached through the lab's own sign-ins, and Voyage embeddings for the corpus and Slack search.
Cloudflare Tunnel
The public route into lab services, such as the subscribe form and the corpus server; any other address gets 'not found'.
12
mintresearch.org top
0%
↓ 0 tokens