Roadmap
The project's goals are written as obstacles: what stands between us and the goal. Each has an owner, an exit (what counts as taking it down) and a date. Verification, not the owner, decides when an exit is met.
Dates are compared with October 22, 2026, the day of the latest change. No obstacle is past its date.
Timeline
One row per obstacle. Bar: from today to the exit date. Square: exit. Diamond: intermediate milestone. Red: date passed and not closed. Click a row to open its file.
All obstacles
| # | Obstacle | Owner | Exit | Date | Status |
|---|---|---|---|---|---|
| V4 | 13 procedure rules undecided Charter: done (D120, adopted at session 1 close, c24b135). A50 confirms. |
A50 -> chair | Decided in session 1 (T7-A50-2) | October 20, 2026 | Exit met |
| O13 | Results that can't be reproduced After month 1 the exit is ongoing: every sealed prediction re-runs before sealing. |
A10 | Run manifests + one seeded re-run in CI (month 1); then every sealed prediction and published number re-runs from its manifest before sealing | November 20, 2026 | Open |
| O14 | Gates that rely on trust (shared token, D68 refused) Evidence = M2 batch merge hash + first green drift-job run. Date set by A10 (month 1). |
A10 | M2 batch merged; drift job live with the president's read-only token; seal covers workflows and scripts | November 20, 2026 | Open |
| O15 | Data too big for git Switch check to the P1 file-size job once it lands. |
A10 | Every T5 source with known access fetched by manifest; no committed file > 10 MB outside the allow-list | November 20, 2026 | Open |
| O16 | No public route to SLAM longitudinal data | A2 | A confirmed route with licence (A9/A50), or a written request sent to the NIA programme with metric v0's plan attached | November 20, 2026 | Open |
| O9 | Data access | A9 + A50 | Every T5 source fetched by manifest with its licence recorded | November 20, 2026 | Open |
| V2 | Hidden keys (planted errors, adversarial cases, held-out mice) not isolated Judged by A1, not A50 (D55). |
A50 -> president | Keys in a GitHub Actions environment that needs the president's approval to read | November 20, 2026 | Open |
| O11 | Published lifespan effects are biased (sick controls, unpublished nulls) | A2 | L1 complete with T1-A2-3 flags; share of flagged results reported | January 20, 2027 | Open |
| O12 | Causal links asserted without refutation tests T8 said 'before sim v0 calibration'; charter part 9 dates it month 3. |
A2 | L3 row with its refuting experiment for every sim v0 edge and mapping, approved by A50 | January 20, 2027 | Open |
| V1 | No blind test exists yet: every score so far is retrospective (T1-A50-3) Judged by A1, not A50 (D55). |
A50 + A28 | >= 10 sealed forecasts on not-yet-public results, criteria approved | January 20, 2027 | Open |
| O1 | No measure that reads ageing reliably | A28 + A50 | Metric v0 passes T1 tests 1-4 on public data; tests 5-6 reported | April 20, 2027 | Open |
| O1 | 'A12 blood counts: no right-way control (blind-spot under O1)' | A28 | An intervention that corrects an aged complete blood count is found (dietary restriction itself induces lymphopenia, moving it "older"; old B6 marrow repopulates better than young); closed individually when a right-way control is found | April 20, 2027 | Open |
| O1 | 'A15 fat mass: no right-way control (blind-spot under O1)' | A28 | A late-life intervention acting on the non-monotonic fat-mass trajectory is found (A15 is non-monotonic, so lower fat in an old mouse is not a pass); closed individually when a right-way control is found | April 20, 2027 | Open |
| O1 | 'A8 lung function: no right-way control (blind-spot under O1)' | A28 | A right-way aged-mouse lung readout control is found (compliance rises with age, so "restored compliance" is not a pass); closed individually when a right-way control is found | April 20, 2027 | Open |
| O1 | 'A9 kidney GFR/albuminuria: no mouse right-way control (blind-spot under O1)' | A28 | A mouse intervention that lowers aged albuminuria or raises GFR is found (right-way evidence is rat-only; 50% CR in B6 moved histology, not GFR/albuminuria); closed individually when a right-way control is found | April 20, 2027 | Open |
| O1 | 'H11 plasma inflammation (IL-6/TNF): no right-way control (blind-spot under O1)' | A28 | An intervention that lowers aged plasma IL-6/TNF is found (only tissue effects found - liver p65, ex-vivo macrophage; plasma readout unmet); closed individually when a right-way control is found | April 20, 2027 | Open |
| O6 | Strain and sex dependence Milestone, February 20, 2027: Sex transfer stratum sealed Dated by owners in G160. Strain transfer before any claim stays ongoing after the exit date. |
A28 + A50 | Metric v0 scorecard reported per sex; strain transfer tested before any claim (ongoing after this date) | April 20, 2027 | Open |
| O8 | No mice, lab or budget | A1 | First request sent with O1's scorecard attached | April 20, 2027 | Open |
| V3 | Scorecards checked only by re-running the same code Milestone, December 20, 2026: recompute plan + 2nd-implementation spec Dated by A50 (PR #51, 161772a). Judged by A1, not A50 (D55). |
A50 | Independent recompute of tests 2 and 4 for metric v0 (T1-A50-4) | April 20, 2027 | Open |
| O2 | Causes not separated from consequences Milestone, February 20, 2027: Sim v0 calibrated on ITP controls; null model (T2-A50-1) and test strata (T2-A28-6) sealed |
A28 + A2 | Model's in-silico interventions match >= 80% of test-set effects in direction and lie within the noise ceiling in size | July 20, 2027 | Open |
| O3 | Cancer ceiling Milestone, April 20, 2027: Necropsy cause-of-death table (T4-A2-1) loaded; model's cause-of-death output compared on controls |
A28 (module), A2 (evidence) | Model reproduces causes of death at necropsy in control mice | July 20, 2027 | Open |
| O5 | 3 years per lifespan test Milestone, March 20, 2027: First early-reading -> remaining-lifespan score on held-out mice from public longitudinal data |
A28 | A metric reading at 6-12 months predicts final lifespan in held-out mice (T1 test 4, read early) | July 20, 2027 | Open |
| O7 | Combined toxicity Milestone, April 20, 2027: Adverse findings and worsened axes for every O4 combination Dated by owners in G160. Co-owners A28 (module) + A2 (evidence). |
A28 + A2 | Toxicity check for every candidate combination before a request | July 20, 2027 | Open |
| O10 | Long-lived species unexplained Milestone, April 20, 2027: First L3 rows on naked mole-rat and bat cancer resistance, with the T1 axes they touch |
A2 + A28 | Comparative module lists which T1 axes differ in naked mole-rat and bats | October 20, 2027 | Open |
| O4 | Combinations don't add up Milestone, April 20, 2027: List of published combination studies (L1) with sealed sim forecasts for each |
A28 | Model predicts published combination studies (e.g. rapamycin + acarbose) within their intervals | October 20, 2027 | Open |