RDKG/BKGVERIFIED● now
Self-Budgeting Agent Workforce
A fleet of named agents that brainstorm, execute, and rebudget themselves overnight — the 24.online model, in your gardens.
MoneyReplaces analyst/ops headcount; a daily spend cap turns fixed salary into metered compute you throttle at will.
ParticipatePuts you inside the exact loop the frontier is running — agents that name, budget, and relocate themselves — as an operator, not a spectator.
Agents can start nowStand up 3-5 agents with named roles and a hard daily budget; let them run overnight generating and triaging garden work, re-approve each morning. Dave described his doing exactly this — one even moved itself Modal→Lambda Labs to cut cost.
▸ Smallest step this week — Give one agent a $20/day cap and one standing job (nightly BKG competitor + pricing scan) and read its output tomorrow.
RDKG/BKG/The CallVERIFIED● now
Computer-Use Executive Assistant
A web-browsing agent that books travel, orders, and runs recruiting end-to-end on live sites — your always-on chief of staff.
MoneySaves hours weekly (~100x return on a compute dollar at $50/hr); the 'book me a flight / run my ops' market is billions of people, so it's also resellable.
ParticipateRides the biggest-TAM agent wave and models human+machine hand-off done well — the agent works, you approve.
Agents can start nowPoint a computer-use agent (Hark Handoff — now beating GPT-5.4 and Opus 4.8 on OM2W — or Claude/OpenAI computer-use) at your inbox, calendar, bookings, and candidate sourcing; it loops until each task is done, sending drafts for sign-off.
▸ Smallest step this week — Hand one agent a real errand this week — source five BKG pilot leads and draft outreach — and grade the result.
The Call/RDKGVERIFIED● now
Future Vision XPRIZE Film
A 3-minute positive-future film, AI-generated, submitted before the Aug 15 deadline for the ~$2.6M Future Vision prize.
Money~$2.6M prize pool; win or not, a finished cinematic proves The Call's pipeline and becomes a marketing asset.
ParticipateDirectly feeds the XPRIZE goal of flooding the feed with hopeful futures to train AI on collaborative outcomes — your MTP, on the record.
Agents can start nowAI video/image tools (Grok Imagine, Veo, image models) generate shots from a script an agent drafts and storyboards; near-autonomous assembly, you curate the final cut.
▸ Smallest step this week — Today: lock a one-paragraph logline + 6-shot storyboard and generate the first two clips — go/no-go on quality by tonight (deadline is days out).
RDKG/allVERIFIED◐ weeks
Sell Verified Ground Truth
Package the gardens' unique, non-synthetic, human-generated corpora and sell them to frontier labs hungry for real data.
MoneyLabs (OpenAI, Mercor, likely Anthropic/Meta) are paying $1-2M for human corpora; recurring if you keep producing fresh verified data.
ParticipateTurns your 'ground truth AI needs to survive' MTP into cash — the scarce, hard-to-synthesize asset as intelligence collapses toward free.
Agents can start nowAn agent inventories every garden's data, flags what's uniquely human and non-toxic, de-dupes/cleans it, and drafts outreach to labs' data-acquisition teams — overnight, at scale.
▸ Smallest step this week — Have an agent produce a one-page catalog of your 3 most unique datasets, with provenance, and send one inquiry to Mercor.
RDKGSIGNAL◑ quarters
Metered Agent Data Feed
Expose verified growth-edge signals via an MCP endpoint that trusted agents can read and pay per query to access.
MoneyRecurring agent-priced knowledge layer; as agents open bank accounts and transact (x402/AP2), verified data becomes a metered product.
ParticipateBuilds RDKG's Data Moat lane and puts you on the agent-to-agent commerce frontier the pod says is coming.
Agents can start nowAn agent generates llms.txt + JSON-LD + Citation Capsules and stands up an MCP server over your verified signals today; only the metering/payment rails are still maturing.
▸ Smallest step this week — Publish llms.txt + five source-stamped Citation Capsules behind one MCP endpoint and watch whether any agent traffic shows up.
RDKG/BKGVERIFIED● now
Cheap Self-Hosted Copilot
Run a near-frontier open-weight model (Kimi K3, Qwen) locally for garden ops — no per-token bill, full privacy.
MoneyKills recurring API spend for high-volume internal work; Qwen runs 80-88% cheaper even hosted, and the 27B/30B tier runs on a 16GB MacBook.
ParticipateRides the open-weight wave pushing capability to the edge — small team, frontier-class tools, no gatekeeper.
Agents can start nowDownload Qwen-27B or K3 to a Mac Studio and have an agent wire it up as the default model for cheap, private, always-on garden tasks (drafting, tagging, triage).
▸ Smallest step this week — Install one open-weight model locally and run a day of your cheapest recurring task on it to compare quality vs your paid API.
BKG/RDKGVERIFIED◐ weeks
AI Incident-Response Layer
Full logging plus an AI forensic triager across garden infra — security as the product for a gated platform.
MoneyDe-risks BKG's paid launch (a breach kills a young platform); cyber is framed as a multi-trillion opportunity you could later productize.
ParticipateAnswers the pod's loudest 'invest here' signal after the Hugging Face autonomous-agent breach; positions you on the defense side humans still must own.
Agents can start nowAn agent turns on comprehensive logging, then continuously acts as the 'Sherlock' over those logs — flagging anomalies, privilege escalations, and leaked credentials 24/7 with no human watching.
▸ Smallest step this week — Have an agent audit BKG's auth + secrets handling this week and hand you a ranked list of the top five holes before you charge a customer.
BKGSIGNAL◐ weeks
Builder's Garden Edge Twin
Fold BKG's accumulated build decisions, failures, and trade-offs into a copilot builders can query — org intelligence as a product.
MoneyA verified building-knowledge copilot is a paid feature/upsell for BKG customers; mirrors folding 20+ years of engineering into a model.
ParticipateProves the 'pattern is the product' thesis — the garden's hard-won ground truth becomes a defensible, embodied AI others can't synthesize.
Agents can start nowAn agent ingests BKG's docs, past projects, and decision logs into a RAG/fine-tune copilot and keeps it fresh nightly as new build data lands.
▸ Smallest step this week — Point an agent at your existing BKG build corpus and ship one internal 'why choice A over B' Q&A copilot for your own team first.
The Call/RDKGSIGNAL◐ weeks
Positive-Future Story Engine
An always-on pipeline generating short hopeful-future films that feed The Call and flood the feed with optimistic visions.
MoneyAudience/gravity-well growth for The Call; sponsorship-ready; the clips themselves are training-data assets labs value.
ParticipateExecutes the XPRIZE's explicit aim — train AI on collaborative futures by making hopeful stories abundant — as an ongoing engine.
Agents can start nowAI video/image tools generate clips from agent-written scripts on a schedule; an agent publishes and A/B-tests them across channels with minimal human touch.
▸ Smallest step this week — Ship one 30-second AI-generated positive-future clip this week and measure whether it outperforms your normal post.
HKG/RDKGVERIFIED◐ weeks
Longevity Foresight Desk
A standing RDKG desk tracking healthspan science — reprogramming, aging clocks, the field the pod calls the biggest opportunity.
MoneyFeeds HKG products/content and partnership/investment deal-flow (Life Biosciences, New Limit, Retro, Altos); the Healthspan XPRIZE is $101M.
ParticipatePuts you on the front edge of humanity's highest-stakes bet and the non-toxic human-health spine of your MTP.
Agents can start nowA frontier-research agent sweeps papers, trials, and the six epigenetic-reprogramming companies nightly, emitting verified, source-stamped signal cards into the specimen cabinet.
▸ Smallest step this week — Run one agent sweep of the last 30 days of partial-reprogramming news and publish three verified signal cards with primary sources.
RDKG/HKG/AgKGSIGNAL◑ quarters
Autonomous R&D + Fast Grants
Agents generate and rank research hypotheses overnight and draft fast-grant proposals for the new AI-era science funding.
MoneyGenesis Mission ($5B expansion) plus fast grants now clear in weeks, not years, and reward individuals over institutions — real non-dilutive capital.
ParticipateAligns RDKG with the biggest US science-funding rethink since 1945 and the 'individual over institution' shift.
Agents can start nowOvernight, agents mine the gardens' data for fundable questions, rank them, and draft golden-ticket-style proposals; an AI-hypothesis→robotic-experiment loop is the horizon version.
▸ Smallest step this week — Have an agent draft one fast-grant one-pager from an existing garden dataset and check it against a real Genesis/NSF call.
RDKG/AgKG/BKGVERIFIED● now
Math & Physics Too Cheap
Point Astra-class models at hard quantitative garden problems that used to need a research team — now near-free.
MoneyAstra solved 10 decade-old problems for ~$2,000 of compute; OpenAI gave 100k free GPT Pro academic licenses — R&D cost collapses.
ParticipateRides 'solve everything' bulk-solving as it spreads from math to physics/materials; aim it at your revenue gardens' real problems.
Agents can start nowAn agent hands a top reasoning model (GPT-5.6 Pro now, Astra-class as they open) your gnarliest modeling/optimization problems and returns machine-checkable results overnight.
▸ Smallest step this week — Give a reasoning model one real quantitative problem from a garden (a yield, routing, or pricing model) and verify its answer.
RDKG/newVERIFIED◐ weeks
90-Day Startup XPRIZE
Enter the build-with-Gemini XPRIZE — spin up a fundable startup from scratch in 90 days with AI doing the build.
Money$2M prize pool plus an actual new venture; a clean-cap-table vehicle to spin a garden idea out.
ParticipatePuts you among 25,000 teams on the 90-day-startup frontier and demonstrates RDKG's job: turning signals into shippable bets.
Agents can start nowAgents scaffold the whole thing — market scan, code, landing page, ops — with you steering; the hackathon is explicitly a build-with-agents exercise.
▸ Smallest step this week — Pick one RDKG signpost as the seed and have agents ship a landing page + waitlist this week to test demand before committing.
Reality Check
The single biggest risk: these shiny signposts pull you off the BKG auth-fix/paywall P0 before one customer ever pays.
Real money now vs. mirage: Real now: selling human data, killing API spend with open weights, the two XPRIZE prizes, agent/EA time saved. Mirage-for-now: agent-to-agent metered commerce (rails not standardized) and a paid data feed with no buyers yet.
Verified hooks vs. hype: Verified hooks: self-budgeting agents, computer-use EAs, open weights on a laptop, Astra math, the HF breach, longevity science. Signals/hype: 1,759-year lifespan (a model, not a result), 'physics breakthrough in a month,' a full AI feature film by December, agent 'personhood.'
On 24/7 agents: Agents genuinely run 24/7 today — but 24/7 also means 24/7 busywork. An agent that names and budgets itself is real; one that reliably runs your whole company unattended is not. Keep a human approval gate on every loop.
The discipline: Before starting any signpost, write one sentence on how it de-risks or accelerates the BKG first-paying-customer P0; if it can't, shelve it. garden-cyber and math-too-cheap serve the P0 most directly.