£1k token sprint

Spend compute like a machine that wants receipts.

Future Union can use agents to generate ideas, rank missions, execute the grind and report progress. The scoreboard decides what counts before the tokens start burning.

EOS-style scoreboard

What the machine is measuring.

JSON feed data_object

Compute budget

£1k

Rules ready

Target: Spend only through ranked missions

Useful outputs

1

Seed controller output

Target: 10 accepted outputs

Review backlog

0

Clear

Target: Never exceed reviewer capacity

Top bottleneck

Direct channel

Needs QA

Target: Signup, welcome, unsubscribe verified

Public proof

Pending

Needs accepted mission output

Target: One weekly machine log

Budget rule

Generate. Rank. Execute. Review. Package.

The point is not to spend tokens. The point is to convert compute into useful, reviewable public work that compounds the next mission.

20%

Generate

Produce aligned mission ideas, blocker hypotheses, source trails, and outcome concepts.

Idea batch with sourceability notes

15%

Rank

Score ideas by constraint, leverage, reviewability, time-to-value, and risk.

Ranked mission queue

45%

Execute

Run top missions into source packs, blocker cards, Builder File updates, Growth inputs, or site/tool improvements.

Reviewable output

15%

Review

Check source trails, caveats, privacy/legal risk, public usefulness, and human ownership.

Pass, hold, or rework decision

5%

Package

Turn reviewed work into public scorecard updates, launch copy, share cards, and proof artifacts.

Public machine update

Ranked queue

The optimal order, until the bottleneck moves.

The queue is deliberately provisional. After each reviewed output, the machine should rerank against the next constraint.

01
Review Direct channel verification

Supporter Channel QA

Broad launch and heavy compute should not outrun the channel that catches committed supporters.

Output: Signup, consent, welcome, unsubscribe and role-tag QA note.

Score

26
02
Generate First local proof

Send One Local Blocker

Local proof is the easiest public first action and feeds receipts, Journal, and Growth Index.

Output: A raw blocker lead with public source, caveat, and review route.

Score

24
03
Generate Source supply

Send One Source Trail

Sources are the fuel for receipts, Builder Files, Journal leads, and Growth Index inputs.

Output: A public source trail ready for receipt review.

Score

23
04
Execute Trust engine

Receipts Desk

The machine only earns trust when claims become inspectable receipts.

Output: A narrow claim card with source, checked date, confidence, caveat, and human owner.

Score

22
05
Execute Ally-facing credibility

Builder Files Source Pass

Builder Files should feel useful and fair before Future Union asks useful groups to engage.

Output: One Builder File upgraded with public receipts, caveats, current ask and suggested mission.

Score

20
06
Execute Positive scene intelligence

Britain Build Radar

Future Union should find useful people and blocked work before the public story becomes only decline.

Output: Five sourced builder/blocker leads with recommended follow-up mission.

Score

19
07
Execute Concrete blocker evidence

Planning Blockers Sprint

High-value proof, but it carries more review burden than simple local blocker and source-trail inputs.

Output: Ten documented planning-delay examples with source trail and confidence notes.

Score

17
08
Execute Method credibility

Growth Index v1

Important flagship work, but it should follow reviewable source supply rather than become a fake leaderboard.

Output: Source pass and method note for one Growth Index input.

Score

16
09
Generate Structured local inputs

Issue Map Intake

Useful after the first blocker/source route is working and reviewers can triage submissions.

Output: A small issue map with facts separated from opinion and source availability flagged.

Score

14
10
Rank Reranking after evidence

Compute Sprint Controller

The controller should rerun after direct-channel QA and the first local proof outputs, not before.

Output: Updated mission queue after the next bottleneck moves.

Score

12

Ranking formula 

priority_score = launch_gate_weight + downstream_unblock_weight + time_to_value + sourceability + user_visible_impact + agent_suitability - legal_or_privacy_risk - human_review_burden - dependency_blockers

Top mission now: Supporter Channel QA. This should change when review data proves a different bottleneck.

Review gate

  1. What changed?
  2. What source trail supports it?
  3. What caveat prevents overclaiming?
  4. Who owns the public claim?
  5. What should happen next?