The 2025 roadmap, scored: 22 promises, 5 kept, 9 dropped

The 2025 roadmap still sits on this site in the future tense. "We're entering a pivotal year," it says, and then lists four bets and five chairs, 22 promises in all. It is October 2026. I went through every promise, looked for what happened to it, and scored it kept, half kept or dropped.

The tally is 5 kept, 8 half kept and 9 dropped. Every score has its evidence in the table below. Where the records cannot settle a score, the row says what I could not find.

1TL;DR

All five promises we kept became a page, a policy or a small script: two service pages, a delivery report, the hybrid-working policy and a Discord tipping flow. Five of the nine we dropped needed someone's weekly hours: an apprentice batch, a bigger delivery team, translations, partner workshops and a VC channel. The other four had other causes. Spatial computing never found demand. The chair model got replaced. The labs treasury never got a budget line, and the DFG earning design never got adopted.

The counts point the same way. Contractor hours logged peaked at 5,458 in March 2026 and fell 68% by September. Pull requests merged across the four vault repos fell 56% from 2024 to 2025. The last case study on this site is dated August 2025.

2What we said in January

The plan has two layers. Four bets: AI engineering and agentic systems, blockchain, platform engineering and spatial computing. Five chairs, each with its own initiatives. Learning promised a labs treasury, build-logs, an AI apprenticeship batch and AI as a copilot for everyone. Delivery asked three questions about growth, quality and measurement. Communication promised short video, Vietnamese translations and an IP character. Partnership promised WALA workshops, VC packages and retainers that lead to platform ops. Engagement promised hybrid working, a gossip protocol, DFG earning by contribution weight and a recognition framework. One more line promised to "formalize our chair-based leadership model."

That makes 22 promises: 21 initiatives and bets, plus the chair model itself. The pages carry two dates. The chair pages say 14 January 2025. The roadmap says 15 June 2025, and the archive shows the roadmap and all five chairs landing in one commit that day. The anniversary post Ten years strong had already named the four bets and five chairs in January. I scored against the year.

I scored a promise kept when the thing exists and works today, half kept when it exists but thinner, later or unmeasured, and dropped when I found no trace or an explicit replacement. The evidence comes from pages on this site, public repos, our Notion databases and the vault's own history. Where the records hold people data I used counts only.

3The scorecard

Hybrid working scores kept because it works, though it was policy before the roadmap named it. Take it out and the tally is four kept.

4What the numbers say

4.1The delivery team shrank

The Delivery chair's first question was how to grow the team. I can measure the answer from December 2025, when every contractor started filing hours through a progress form.

Logged hours rose from 4,456 in December to 5,458 in March, then fell every month after March. September logged 1,728 hours, 68% below the peak, from 13 contractors against 35 in March. The roster says the same. Our Notion database held 27 active contractors on 21 July, 24 on 27 August and 18 on 18 October.

The dated rows for earlier months are thin. Of 275 contractor rows, 114 carry a start date and 57 of the 257 inactive rows carry an end date. Those rows show 7 starts and 9 ends dated in 2025, and 3 starts and 27 ends so far in 2026. The dates are incomplete, and the direction is still not growth.

Two causes explain most of the decline: client roll-offs through summer 2026, and AI efficiency shrinking the billing per person.

4.2Output fell after 2024

The Communication chair wanted content to build an audience. Output fell in 2025.

Across research, brainery, handbook and playbook, 333 PRs merged in 2024 and 148 in 2025. Case studies went the other way before they stopped: 3 dated in 2024, 5 in 2025 and none in 2026. The last one, on monitoring our own ICY backend, is dated 7 August 2025.

The Learning chair promised a build-log for every experiment. The vault holds 77 build-log entries dated 2025, and 62 of them carry a date in May. Ten stand alone and 67 sit in eight series. Today 25 of the 77 are live, 10 are drafts and 42 are gone, cut in the two August 2026 triages. I matched the survivors by title, so treat the split as close but not exact.

4.3The pipeline we cannot score

The Partnership chair listed success metrics with no numbers: "Maintain efficient sales cycle for WALA engagements" and "Establish partnerships with multiple VC firms". The deals database could not have scored them anyway. It holds 28 deals. Nineteen were created between July 2022 and February 2023, seven have no create date, and one each was created in December 2025 and April 2026. In June 2026 one prospect was open. Nothing I could read records the 2025 pipeline.

5What happened to the three chairs we deleted

The delivery, communication and engagement pages are gone. brainery PR 214 deleted them on 17 August 2026, among 609 files cut in one triage of low-value content. Its description lists them as "generic chair descriptions". The learning and partnership pages survived the same triage. I recovered the deleted three from the vault's history and read them against the roadmap.

All three ended with success metrics that carried no number: "Increase client satisfaction scores", "Grow total social media following significantly" and "Achieve high team satisfaction score in quarterly surveys". A promise with no number cannot be scored, so I scored the initiatives the roadmap page names. Here is where the chairs landed.

The triage kept the two pages that named specific programs. Those two scored worst: none of their seven promises counts as kept. The three deleted pages hold three of the five kept promises. Specifics on the page did not predict delivery. The roadmap page and the two surviving chair pages are drafts now, and their URLs redirect here.

6Why things stopped

This section is my reading of the records.

The pipeline went quiet. The bets that needed outside demand had nothing to hang on. Spatial computing, VC packages, WALA workshops and retainers all depend on a conversation with a buyer. The deals database holds no 2025 row before December, and the live homepage sells an assessment, hardening and Run and Own to finance and technology companies. It has no spatial offer and no VC package.

AI changed the junior math. The apprentice program planned six months on live client projects, with applications opening in late 2025. They never opened. A draft we wrote in September 2026 for a replacement reasons the other way: AI removes the paid junior work, so it removes the reps juniors used to get, and a program has to build those reps on purpose.

A weekly cadence needs a name. Video, translations and WALA all needed a person's hours every week. The roadmap named a chair for each and no person. Gossip tipping survived because ops reviews each tip and pays it from the vault.

7What 2026 inherited

The chair names gave way to nine role manuals in April 2026, written for a firm that works with AI agents. The apprentice batch became a drafted senior-pipeline program in September. The build-log habit became field reports with measured tables, such as Memo, rebuilt and the deploy pipeline formula. The positioning moved from four bets to a plain offer: assess a system, harden it, and run it.

8What to change next

A promise goes on the page with a number and a date, or it stays in the notes. A promise that needs weekly hours names the person who gives them. The next roadmap should carry both, plus a re-score date on the page itself, so the reader can see how the last one did.

9Limits

• The scores are mine. Nine rows rest on judgement the records cannot settle: blockchain, the chair model, the labs treasury, the apprentice batch, video, the IP character, VC packages, retainers and DFG earning.
• Hours are estimates. The progress form captures approximate effort per entry, and it went live in December 2025, so I cannot compare 2025's first half.
• PR counts count PRs. They do not count words, and 2026's brainery PRs mostly restructure the vault.
• I could not read channel analytics. The video row rests on a write-up and the absence of links.
• I searched names, pages and databases. Private client work or code under a different name could hold spatial or blockchain work I did not find.

10Where the three notes went

The three notes are now drafts, and their URLs redirect here, along with the URLs of the three chair pages deleted in August 2026.