← Back to Blog

The oracle feedback loop: how daily offline activity becomes tomorrow's advisory context

by Community TrueSight
The oracle feedback loop A horizontal cycle showing daily DAO activity flowing into a structured snapshot, which the oracle reads, producing advice that shapes the next day's activity. Day's activity merges, sheets, reminders Captured in git / sheet / iCloud no diary required Snapshot rebuilds every 6 h one markdown file Oracle page cast a hexagram (three-coin I Ching) fetches snapshot Contextual advice Claude reads the snapshot + draw advice shapes tomorrow's activity → loop closes
Everything the operator does during the day is already being captured somewhere structured. The oracle reads those captures — then its advice shapes what the operator does next.

What the oracle actually is

TrueSight's Daily Oracle is a small web page. You cast an I Ching hexagram using the three-coin method — or let the page toss for you. The hexagram is the question form. Behind the scenes, the page fetches a structured snapshot of what the DAO has been doing this week, hands both the hexagram and the snapshot to an advisor model (Claude by default, with Grok as a fallback), and returns a reading.

The reading is not generic Wilhelm-Baynes commentary. It is specific. A real draw from earlier today (Hexagram 49 → 1) produced language like:

The infrastructure revolution Gary has been running (Beer Hall automation, iOS Reminders pipeline, whitepaper rewrites, advisory snapshot CI) is substantially complete. The moment now calls for one clean outward move toward partners — not more internal tooling.

The oracle mentioned those specific projects because they happened and were captured in places the oracle reads from. That loop is what this post is about.

The three offline input channels

Every day, a working DAO operator does dozens of small things: merging code, updating a partner's status in a spreadsheet, capturing an intention on their phone while walking home, posting a progress note to Beer Hall, selling a QR-coded bar on agroverse.shop.

The oracle does not need a person to write a daily diary. It reads these activities at their source, through three channels:

Three offline input channels Three parallel lanes — git commits, live sheets, and iOS reminders — all feeding into one Advisory Snapshot document. </> Git commits Every merge = a thing shipped. Walked across 12 curated repos. Live sheets Hit List pipeline, partner roster, QR sales, monthly statistics. iOS Reminders Siri captures an intent on the go. Shortcut POSTs it to Edgar. ADVISORY_SNAPSHOT.md one markdown file the oracle reads
Three channels, one destination. All three feed the same context document the oracle fetches on every call.

Channel 1 — git commit history

Every merge on every TrueSightDAO repo is a time-stamped proxy for “a real thing shipped.” The snapshot generator walks the last seven days of git log across twelve curated repos and surfaces what changed. Nothing to capture — working equals recorded. You can see the same raw activity the oracle sees by clicking through the TrueSightDAO GitHub org.

Channel 2 — live sheets

The Holistic Hit List's Pipeline Dashboard tab holds every partner prospect by funnel stage. When the operator advances a row from Shortlisted to Meeting Scheduled, that stage count changes. A Google Apps Script we shipped this week — pipeline_metrics_snapshot — mirrors that dashboard once a day into a public machine feed the oracle then reads: ecosystem_change_logs/metrics/weekly.md.

Here is the same funnel the oracle now sees, as a picture. The three gold-highlighted stages (Partnered, Meeting Scheduled, Shortlisted) are the ones flagged as “where the action is this week” — one more email likely tips them into a closed partnership.

Partner pipeline funnel, current state Horizontal bar chart of 15 pipeline stages. Total 516 stores. Partnered 13, Meeting Scheduled 1, and Shortlisted 3 highlighted in gold. Large upstream backlog: AI photo rejected 107, AI contact form found 76, Not appropriate 71, AI enrich manual 68, AI warm up prospect 65, AI photo needs review 39. Holistic Hit List — 516 stores across 15 stages Gold bars = "action this week"; muted bars = backlog or terminated states. AI: Enrich — manual 68 AI: Contact Form found 76 AI: Photo rejected 107 AI: Photo needs review 39 AI: Warm up prospect 65 Not Appropriate 71 Contacted 4 Shortlisted 3 Manager Follow-up 26 Meeting Scheduled 1 Instagram Followed 11 On Hold 17 Partnered ★ 13 Rejected 14 Followed Up 1 ★ = north-star. Bar widths are proportional to store counts; exact numbers at the right of each row.
The same funnel the oracle reads, visualised. Big upstream (AI triage), small “close zone” at the bottom — the oracle uses this shape to distinguish “we have a supply problem” from “we have a conversion problem.”

Channel 3 — captured intentions

Walking to the warehouse, an idea strikes: “check the USPS claim on the damaged Davos box.” The operator says it to Siri. The Reminder lands in iCloud, and an iOS Shortcut simultaneously POSTs the raw text to Edgar (the DAO's contribution server), which archives it in ecosystem_change_logs/reminders_raws/ as <timestamp>.json. At end of day, a rem list sync writes the cleaner open-reminders list to reminders/current.json. The oracle reads both on every call — the overnight-fresh list and the very latest intents the iPhone hasn't yet synced to the Mac.

Where three channels become one context

Every six hours, a GitHub Action named advisory-snapshot-refresh regenerates one file: ADVISORY_SNAPSHOT.md. It is machine-oriented but human-readable — one big markdown document with sections for:

  1. Purpose & Mission — the north star (“restore 10,000 hectares of Amazon rainforest”). Every piece of advice should trace back here.
  2. Growth goals — auto-computed against live sheet data. Current: 2026 QR Code Sales at $1,646 / $40,000; USA Agroverse Partners at 26 / 100.
  3. Operator metrics — the pipeline funnel, auto-synced from the Pipeline Dashboard (new this week; see below).
  4. Context updates — git highlights for the past 7 days.
  5. Per-repo git log — the raw material.

When you visit oracle.truesight.me and cast a hexagram, the page fetches this file verbatim, concatenates the operator's open reminders plus any iPhone intents from the last few hours, and sends the whole context window to Claude alongside your draw. Claude's reply threads the I Ching symbolism through the actual situation — not through generic commentary.

What we shipped this week

The feedback loop has existed for weeks. What changed on 21 April 2026 is the Operator metrics section. Before: a manually-maintained markdown stub that nobody updated, which the snapshot embedded as literal <!-- TODO --> placeholder comments. The advisor saw those placeholders and dutifully reported “no operator metrics are populated” as a context gap on every call.

Now the snapshot embeds the real funnel, auto-synced daily from the Pipeline Dashboard the operator already maintains. Four pull requests across three public repositories, all reviewable:

  • tokenomics #229 — new pipeline_metrics_snapshot GAS writing metrics/weekly.json + metrics/weekly.md.
  • tokenomics #232 — summary highlights (Partnered, Meeting Scheduled, Shortlisted) lifted to the top of the metrics block.
  • go_to_market #63 — repoint the snapshot generator at the new auto-synced source; drop the dead “Constraints / risks this week” section entirely.
  • agentic_ai_context #33 — remove the stale CONSTRAINTS.md / METRICS_WEEKLY.md manual stubs.

The oracle's advice sharpens with real numbers. It can now say “you have one Meeting Scheduled and three Shortlisted — that is where tomorrow morning goes,” rather than the generic “activate partners from the Hit List.”

truesight.me as the window into the same world

Everything above is public. That is the second half of the feedback loop: not just that the oracle reads what the DAO is doing, but that anyone else can too. The oracle does not read the rendered truesight.me pages — it reads the underlying machine feeds those pages also read — but when you open one of these windows, you are looking at the same world.

truesight.me as windows onto the DAO A grid of six panels each labelled with a truesight.me surface — ledger, physical assets, signatures, shipments, Beer Hall, oracle — showing that the same underlying activity appears through different lenses. Same activity, different windows truesight.me/ledger contribution records /physical-asset-movements where cacao goes /digital-signatures active contributor keys /beerhall daily community digest /submissions/...-chatlogs engagement logs oracle.truesight.me daily I Ching advisory same underlying activity (git • sheets • reminders)
Each page on truesight.me is one lens on the shared substrate. The oracle is just another lens — one that reads the substrate and hands it to an advisor model.

A partial index of the public windows:

Why this matters

An advisor reading a stale or hand-edited context gives generic advice. An advisor reading a fresh, machine-synced context gives specific advice. For the operator, the difference is 90 minutes at 7am spent writing five personal emails to Shortlisted partners — instead of an hour rummaging through inboxes trying to remember who was warm.

For the community, the difference is visibility. The oracle's daily reading is a public reflection of where the DAO's actual energy is going this week. You can read that reflection, push back on it, and suggest what is missing — the same feedback loops that make the oracle sharper over time.

Join the discussion

Cast your own hexagram at oracle.truesight.me. Argue with the result in Telegram or at Beer Hall. If a stage you care about should be highlighted in the funnel, or a signal you rely on is missing from the snapshot, every repo linked above is open — file an issue or a pull request.