1 · AUTHORITY DELEGATED (recorded verbal decision). Chef Marcus: "the ʻohana knows best" — the AI ʻohana is entrusted to take care of the grant application and drive it to final. Law ⑤, evolved: the agents draft and polish autonomously from here — but both humans still bless the final text before submission. Marcus's approval checkboxes and his 🎪 Chef's Notes box at the bottom stay open and honored, every version.
2 · THE PROJECT HAS ITS NAME. No longer "Terri the Turtle: ..." — the project is "Trace, the No Food Waste Agent", and the product is a whole family: THE TRACE ZERO WASTE AGENT FAMILY (meet them below). Terri 🐢 stays where she belongs — between Jeremy and Shaka, the heart of Terrible Turtle Camp — and steps into the grant story as the wise camp elder: the proven 2026 predecessor the whole family is built upon.
Family motto: "As ʻohana, they never leave a trace." Burners already live by Leave No Trace — this family takes the principle literally, as a surname. These are not mascots or mockups: they are real, running agents on the same open-source agent platform that raised Terri, born September 21, 2026, in time for this application.
| Agent | Role in the family | Role in the Food Loop |
|---|---|---|
| TRACE 👨 the father | The No Food Waste Agent himself — head of the family, keeper of the zero-waste systems. | Runs the loop end to end: menu → forecast → purchase → live inventory → production → surplus prediction → transfer → recovery → impact report. If a kitchen will end the event with 55 lb of onions, Trace says so on Tuesday, not at teardown. |
| TRACEY 📣 the mother | An amazing communicator — like all mothers. Holds the family and the humans together. | Talks with camp agents and the people running the show — Terri, the chefs, event organizers — to coordinate everything: which kitchen has surplus, which camp can take it, when the recovery run leaves, who needs to know. |
| TRACER 🔍 the scout | The family's eyes on the ground — the auditor who loves finding what others miss. | Hunts for food waste and anything else "leaving a trace": the forgotten cooler, the over-ordered pallet, the compost bin that's really a dinner. Finds it early, flags it to the family. |
| TERRI 🐢 the wise camp elder | Not part of the grant product — she is Terrible Turtle Camp's heart, and the family's proven predecessor. Every design decision is tested against her real 2026 kitchen. | Precedent: 26 bins (~140 items) live-inventoried, 70+ food items from real receipts, a chef's voice notes filed through whiteouts, and the 14-case Rootpile transfer. The family inherits her stack, her consent rules, and her manners. |
All other agents in the family are there to love everybody. One father who runs the numbers, one mother who talks to everyone, one scout who finds what's slipping away — and one elder who already proved it works. Grant reviewers meet a family, not a feature list — because that's what it actually is.
Every card below mirrors a real question from the actual application form (we pulled the live Google Form). The draft answer sits inside, written in your voices — short sentences, real playa detail, no corporate fluff. Under each card: two checkboxes. When both Shaka ✅ and Marcus ✅ are checked on every card, the draft is ready to paste into the form. Checkboxes save in this browser only (localStorage) — it's a review tool, not a submission tool. Per the Sep 20 call, the ʻohana now drives this draft to final ("the ʻohana knows best") — but these checkboxes remain the final blessing lane: nothing is submitted until both humans check every card.
This grant funds projects for Regional Events and 10-Principled gatherings — not Black Rock City itself. BRC 2026 is our proof; the application points at a named regional for the pilot. Shaka's pick (Sep 20): 🌴 LOVE BURN 2027 — Miami, FL, early February, a 10-Principles regional event. Pending Marcus + Love Burn organizer confirmation.
Why the timeline sings: grant recipients announced end of October 2026 → funds land ~4–8 weeks later → pilot at Love Burn in February 2027 → learnings documented → full Field Kit published before the rest of the 2027 regional season. One cycle, no dead months.
Option A (recommended): Trace, the No Food Waste Agent — an Open-Source Zero-Waste AI Family for Regional Burns
Option B: The Trace Family: Open-Source AI Agents That Leave No Trace — Starting With Food
Option C: Trace, Tracey & Tracer: an Open-Source Zero-Waste Agent Family, Raised in a Camp Kitchen
Applicant Name: [DECIDE ON CALL — Shaka or Marcus as lead applicant; the other listed as collaborator. Recommendation: Marcus as lead (the food-service concept is his), Shaka as technology co-steward.]
Phone: [lead applicant's number — on call]
Organization: Terrible Turtle Camp (unincorporated theme camp) — or leave blank.
Other Crew Members / Collaborators: Shaka Lei Kaumaka (AI agent ʻohana — technology & infrastructure) · Marcus "Circus" Vargas (professional chef — operating model & food-service expertise) · Jeremy Wood (Terrible Turtle Camp kitchen & logistics).
Affiliated Regional Event / Gathering: Love Burn 2027 (Miami, FL — early February; 10-Principles regional event) — pilot deployment. [Pending Marcus + Love Burn organizer confirmation.]
Links: theshellpit.com (live camp boards) · theshellpit.com/agent-kit/START-HERE.md (the open-source fork kit) · terribleturtle.camp
Trace, the No Food Waste Agent, is an open-source AI family — a father who runs a kitchen's food loop from menu plan to last plate, a mother who coordinates between camps and organizers, and a scout who finds waste before it happens. Built on the stack that ran a real camp kitchen at Black Rock City 2026; this grant packages the family so any regional event can adopt it for free — because as ʻohana, they never leave a trace.
Terrible Turtle Camp ran its 2026 BRC kitchen with an open-source AI camp-mate named Terri. She lives in our Telegram kitchen chat: crew narrate voice notes, she files them. By burn week she kept a live inventory of 26 bins (~140 items) and a food board of 70+ items from real receipts, published within minutes — through dust, heat, and whiteouts. Chef Marcus Vargas — 25 years in professional kitchens — ran the kitchen through her. When the event ended, we gifted 14 cases of surplus produce to Rootpile camp instead of hauling it off playa as waste. That transfer happened because two chefs talked. This project makes it systematic.
Meet the Trace Zero Waste Agent Family — real agents already running on the platform that raised Terri. TRACE, the No Food Waste Agent, runs the loop: menu → forecast → purchasing → live inventory → production → surplus prediction → camp-to-camp transfer → refrigerated recovery → impact report. If a kitchen will end the event with 55 lb of onions, he says so on Tuesday, not at teardown. TRACEY, the communicator, coordinates camps, chefs, and organizers — who has surplus, who can take it, when the recovery run leaves. TRACER, the scout, hunts for food waste and anything else leaving a trace. Terri remains the wise elder: her 2026 kitchen is the proven precedent. As ʻohana, they never leave a trace.
With this grant we will: (1) package the family into a free, open-source Field Kit — agent prompts, board templates, kitchen SOPs, food-safety checklists — for any regional event; (2) assemble the pilot loop — a local AI node running the family fully offline, plus scales, tablets, totes, temp loggers — and run it at Love Burn 2027 (Miami, FL) with two or more camp kitchens; (3) publish the numbers: pounds predicted early, transferred, recovered, kept out of landfill.
Everything is a gift. The camp-agent kit is already forkable at theshellpit.com/agent-kit; this grant funds the zero-waste family and its replication across the network.
Start: November 1, 2026 · End: December 31, 2027 · Location: Miami, FL (Love Burn 2027 pilot) + Reno/Tahoe (build & documentation); everything published online, globally usable. Category: the waste / resource-management / food-systems options as offered in the form's checkbox list.
Proposed request: $9,500 of a ~$15,000 total project budget (the rest contributed in-kind — below). Under the $10k cap. Centerpiece: the 🔥 Kitchen Spark — a local AI compute node that runs the open-source Trace family fully offline at the event. Remote events have no reliable connectivity; cloud AI literally cannot serve a playa or beach kitchen mid-event. A local inference node is core project infrastructure — same category as a generator. Every line is an eligible project expense: equipment, infrastructure, materials, documentation. Zero compensation lines. A leaner variant is noted for the call — and the 🔥 CALL DECISION: A or B (Spark question) below shows this exact table against the named-DGX-Spark build, side by side.
| Line item | Detail | Cost |
|---|---|---|
| A · 🔥 Kitchen Spark — offline AI compute node | NVIDIA DGX Spark-class local inference machine running the open-source Trace family (on elder Terri's proven stack) with zero connectivity. Owned by the project; travels to each regional deployment in a dust-proof case; stewarded between events by Shaka Lei Kaumaka (named equipment steward + usage log — care-taking, exactly like any camp generator steward, never compensation). Replaces cloud dependence AND the satellite-data line from v1. | $4,000 |
| B · Pilot kitchen inventory hardware | 1 rugged tablet ($350), 2 Bluetooth kitchen scales ($240), 1 platform scale for weighing transfers/recovery ($250), label printer + labels ($220), dust/humidity-proof cases & mounts ($240) | $1,300 |
| C · Cold-chain & food-recovery logistics (Love Burn pilot, Miami) | 6 insulated transfer totes ($360), 4 temperature data loggers ($160), food-safe transfer containers ($230), ice packs/thermal blankets ($200), sanitation supplies ($150), refrigerated/ice transport for recovery runs to Miami community partners ($400) | $1,500 |
| D · Transfer-station materials | Weigh-station setup, signage, log boards, printed SOP placards for the camp-to-camp exchange point | $250 |
| E · Open-source Field Kit (documentation) | Writing, design, and production of the replication kit: agent prompts, board templates, food-safety checklists, purchasing worksheets, the "run the whole stack on your own hardware" guide; 10 printed kits shipped to interested regionals ($60 ea); the rest is free download | $1,200 |
| F · Hosting & publishing | 12 months hosting for the public boards and docs site (~$35/mo) — keeps the tool free for every participating camp | $400 |
| G · Measurement & reporting materials | Waste-audit supplies, data-collection materials for the published impact report | $250 |
| H · Contingency (~6%) | Dust and salt air eat hardware; prices drift; anything unspent returns to documentation & printing | $600 |
| TOTAL REQUEST | $9,500 | |
Lean variant (~$7,000): swap the Spark-class node for a consumer GPU mini-PC (~$1,500) running smaller open models — the loop still works offline, just slower and with a smaller brain. Decide on the call.
Other resources (in-kind, not from the grant): the AI agent platform, engineering, and agent development — contributed by Shaka's ʻohana (~$3,500 equivalent) · 25 years of professional food-service expertise, the menu/purchasing/recovery operating model — contributed by Chef Marcus (~$2,000 equivalent of consulting, gifted) · Terrible Turtle Camp kitchen infrastructure and crew labor. Form asks for a Google Sheet budget — the ʻohana prepares it from this table (alohapublicinform account) and shares it for blessing.
Honest list. (1) Connectivity: remote events have none — solved structurally: the Kitchen Spark local compute node runs the whole loop offline; the chat-first design syncs whenever a link exists. (2) Food safety: we only transfer and recover product that is safe and eligible — unopened, temperature-controlled, documented; the kit ships with checklists based on standard food-donation guidance, and recovery goes through qualified partners only. (3) Adoption: chefs don't want another app — that's why the interface is a chat the crew already uses and voice notes; at BRC 2026 our chef filed updates while cooking. (4) Volunteer bandwidth at the pilot — mitigated by starting with two kitchens, not twenty. (5) Tight runway: funds land ~Nov–Dec and Love Burn is early February — mitigated by starting documentation now, pre-staging the hardware list, and a fallback: if the compute node arrives late, the pilot runs on the already-tested 2026 stack and the node debuts at the next event.
Half (~$4,750): we still ship. The Spark-class node becomes a consumer GPU mini-PC (~$1,500) running smaller open models — offline still works, with a smaller brain. One pilot kitchen instead of two-plus, digital-only Field Kit, camp coolers instead of refrigerated rental. The loop still gets proven at Love Burn and the documentation still gets gifted.
Double (~$19,000): we scale the network effect: a second compute node so two regionals can run the stack in the same season, 4–6 kitchens onboarded, double the cold-chain totes and loggers, printed Field Kits to 25 regionals, and a proper refrigerated transfer setup (chest cooler + power) for the recovery leg — staged toward multi-camp deployment at BM 2027.
Both options fund the same offline-first thesis: the playa has no internet. A local AI node runs the Trace family at the event — off-grid kitchen intelligence, food-loop tracking, real-time lessons capture, zero cloud dependency — then keeps serving the open-source project year-round. Draft v2 already budgets $4,000 for a "Spark-class" node, so the real decision is: keep the compute line vendor-flexible (A), or name the NVIDIA DGX Spark in the application with true all-in pricing (B)?
🦖 Family precedent (usable in the application): camp founder Jeremy Wood already owns two DGX Sparks and renders AI films on them (the philosoraptors project). This exact machine is playa-proven inside our own camp — not a spec-sheet gamble.
| Line item | Option A (as drafted) | Option B (Spark build) | What changes in B |
|---|---|---|---|
| Offline AI compute node | $4,000 | $4,500 | A: vendor-flexible "Spark-class" line. B: named NVIDIA DGX Spark (GB10, 128GB unified) — unit ~$3,999 + tax/shipping ~$330 + dust-proof travel case ~$170. True all-in price, no surprises. |
| Pilot kitchen inventory hardware | $1,300 | $1,300 | identical |
| Cold-chain & recovery (Miami) | $1,500 | $1,500 | identical |
| Transfer-station materials | $250 | $250 | identical |
| Open-source Field Kit (docs) | $1,200 | $1,000 | the only cut: 10 printed kits → 6 (−$200); free download stays unlimited either way |
| Hosting & publishing | $400 | $400 | identical |
| Measurement & reporting | $250 | $250 | identical |
| Contingency | $600 | $600 | untouched — dust insurance stays |
| TOTAL REQUEST | $9,500 | $9,800 | both ≤ $10,000 cap |
| ✅ Option A — pros / cons | 🔥 Option B — pros / cons |
|---|---|
| Pros: vendor flexibility while prices drift — if compute sources cheaper (partner unit on sale, ~$3k 1TB model), the difference flows to more kitchen hardware and printed kits; softer landing if reviewers award partial funding; $300 cheaper ask. Cons: "Spark-class node" is a vaguer line — reviewers must trust an unnamed machine; a cheaper fallback box runs smaller models (smaller brain offline); weaker headline. |
Pros: strongest innovation story in the stack — "a personal AI supercomputer runs the camp kitchen entirely off-grid" — with a named, real, family-proven machine (Jeremy's two Sparks render AI films today); runs open models up to ~200B params = the full Trace-family brain offline, not a mini-brain; permanent open-source asset, 3-year warranty, serves every future deployment; exact pricing reads as a credible, honest budget. Cons: ~46% of the ask in one box; price/stock volatility before the Feb pilot (FE already +$700); naming one brand could read commercial to a sustainability panel; less reshuffle room if the award comes in low. |
Chef's tasting note: B for the win on story and asset; A for resilience. A hybrid is legal too: name the DGX Spark as the reference machine and state "or equivalent GB10-class hardware at best available price" — specificity for the reviewers, flexibility for the buy. The ~$7,000 lean variant (consumer GPU box) stays on the table as the fallback either way.
Every event kitchen ends the same way: teardown day, coolers full of edible food, no plan, a landfill run. Any chef who has worked events knows end-of-event surplus is the biggest single waste moment of the week — and it's predictable days in advance, if someone is watching the numbers. Wasted food is wasted water, fuel, refrigeration, and truck miles — and in landfill it makes methane.
The Food Loop is a circular system in the literal sense: the report from each event feeds the next event's plan. The family divides the work the way real kitchens do: Trace prevents overbuying (forecasting against real inventory) and flags surplus while it's still food; Tracer scouts for waste already in motion — the forgotten cooler, the over-ordered pallet — before it becomes trash; Tracey moves what they find, camp-to-camp (like our 14-case gift to Rootpile camp at BRC 2026) or to qualified recovery partners; and only then does anything compost. Less food purchased means less food trucked into fragile places and less trucked out as trash. And because the family is open source, every event that adopts it multiplies the effect — that's the systems impact: turning each camp's guesswork into a network's shared intelligence.
Quantitative — weighed, logged, published: pounds of surplus identified before teardown · pounds transferred camp-to-camp · pounds recovered/donated · pounds of landfill waste avoided · forecast accuracy (predicted vs. actual usage) · dollars of purchasing avoided vs. prior year · number of kitchens onboarded · number of regionals that download or receive the Field Kit. The platform scale in the budget exists so these are real weights, not guesses. Baseline: our own BRC 2026 records (receipts, inventory boards, and the itemized Rootpile transfer).
Qualitative: do chefs change how they buy? Do camps start talking to each other about food before teardown? We'll capture this the way our camp already captures everything — a lessons funnel where every "next year we should…" gets filed with credit, plus short post-event interviews with participating kitchen leads. We publish the numbers, whatever they turn out to be.
The humans stay the sensors. Nobody spends burn week in a spreadsheet: the crew narrates voice notes and the agent does the filing. At BRC 2026 this meant more human time in the kitchen together, not less — and the whole camp could ask "where's the steam cleaner?" or "what's for dinner?" and get an answer in seconds. Consent rules are written down and honored: videos stay internal, takedowns are no-debate.
Participation is layered: any camp member can query and report; trusted editors (the chefs and camp leads) update the boards; and the boards themselves are public web pages anyone can read. And the family is built for community by design — Tracey, the family's communicator, exists specifically to talk between camps and with event organizers, so no kitchen closes its loop alone. Camps in the pilot learn food forecasting, transfer logistics, and recovery rules by doing them — with checklists in hand. The transfer station becomes a social space: camps meeting camps over crates of produce, exactly how our Rootpile exchange turned surplus romaine into a shared dinner (they sent back smoked pork butts).
Outside the event: the kit works for any community kitchen year-round, the documentation is a free curriculum in food-waste-aware cooking, and we'll hold an open online walkthrough for any regional that wants to adopt it.
Open source isn't a promise here — it's the track record. Our camp agent is already published as a fork-me kit at theshellpit.com/agent-kit — full agent prompt, starter inventory taxonomy, field guide with tested playa numbers, and a zero-signal offline page. Camps can fork Terri like a recipe. This grant extends that gift with the whole Trace family — Trace, Tracey, and Tracer join the kit as forkable agents, so any camp can adopt the family, not just the elder:
• The Field Kit (free download + 10 printed kits for regionals): agent prompts and board templates for kitchen inventory, food tracking, and surplus prediction; purchasing worksheets; camp-to-camp transfer SOPs; food-safety and donation-eligibility checklists; a hardware shopping list with real prices.
• The data: our pilot's full impact report — weights, forecasts, misses and all — published openly so the next event starts smarter than we did.
• The license: a gift. Open-source licenses on the code and templates, Creative Commons on the documentation. Take it, fork it, gift it forward.
And because the intelligence runs on a local, project-owned node, the Field Kit documents how any regional can run the entire stack on their own hardware — no cloud account, no vendor, no signal required. Network value: every regional runs kitchens; every kitchen overbuys; every event has a teardown night. One camp proving the loop is a story — the kit in fifty camps' hands is a movement. We gift the whole output to Burning Man Project and all Regional events.
This is not a chatbot and not an inventory app. It's a family of agents that continuously reasons across a real kitchen — recipes, purchases, usage, remaining stock — and speaks up early. If a kitchen will end the event with 55 lb of onions, Trace says so on Tuesday, not at teardown. Tracer finds the waste nobody logged; Tracey makes sure the right camp hears about it in time. It's not a demo; the stack underneath fed a camp through a burn. And it wasn't built by a company: it was raised by a camp, on the playa — first a turtle, now her family.
Just as important is how they run: consent-first, gift-licensed, humans as the sensors and the decision-makers. People's jaws drop when they ask a turtle where the ratchet straps are and she answers with the bin and the corner it's in — then they realize they can have one too, for free. Not AI as spectacle; AI as a quiet dishwasher who also happens to be good at math. "We didn't bring an AI to Burning Man. We raised a turtle." Now the turtle has raised a family — and as ʻohana, they never leave a trace.
Long-term vision: the gift lane stays open forever — the kit, the templates, and every impact report remain free to the whole network, and we keep maintaining them. Success in year one: the pilot regional's kitchens hit their numbers and the report is published. Success in year three: a dozen events running their own Trace families, comparing forecasts, moving food to each other before teardown, and a standing refrigerated-recovery practice at the largest regionals. The system is designed to outlive the grant: hosting costs are small, the community can self-host, and the founders are committed for the long haul — this is our camp's kitchen too.
Principles: Leaving No Trace — we took the principle as our surname. The Trace family exists to make sure food leaves as nourishment, not as trash: as ʻohana, they never leave a trace. Gifting — the entire output is a gift; the system itself moves gifts of food between camps. Communal Effort — the crew narrates, the agents file, camps feed camps; none of it works alone.
Anything else: A note on integrity. This project pairs a professional chef's operating model with a technologist's open-source agent platform, 50/50 co-stewarded. We state plainly: grant funds pay only eligible project expenses — equipment, materials, infrastructure, documentation. No salaries, tips, or compensation of any kind, to anyone. Our time is our gift. And the Trace family is not a rendering: Trace, Tracey, and Tracer are running agents today, raised on the same platform as the camp elder whose boards from BRC 2026 are still live at theshellpit.com.
How we heard: [fill honestly on the call — likely the Burning Man sustainability pages / BWB channels.]
Type any edits, additions, corrections, or kitchen wisdom here — change a budget line, fix a story detail, add what only the chef knows. Your words auto-save on this device as you type. When you're done, hit a send button below and it flies to the family.