How to Roadmap with Decisions Rather Than Dates: Will Larson
Table of Contents
The session
| Field | Value |
|---|---|
| Title | How to Roadmap with Decisions Rather Than Dates |
| Series | Anthropic Partner Series |
| Speaker | Will Larson, CTO at Imprint |
| Date | |
| Time | 10:00–11:00 PT (13:00–14:00 ET) |
| Format | Webinar, one hour, recorded |
| Recording | Link within 48 hours of the event |
The pitch: run a roadmap constrained by decisions rather than dates, an approach the page attributes to recent AI capability rather than to planning fashion. Four advertised mechanisms:
- Empower small teams by reducing handoffs, using AI discovery.
- Minimise cross-functional approval gates via interfaces a machine can drive.
- Eliminate open decisions through rapid prototyping and feature flagging.
- Escalate quickly when the other three are insufficient.
What to listen for
The framing is older than the AI claim attached to it. "Decisions, not dates" is the same move as a decision log, a spike, or a set-based concurrent engineering schedule: the schedule is a symptom, and the real constraint is the queue of unresolved questions. What is new in the pitch is the assertion that AI shortens the time from open question to closed one, mostly by making prototypes cheap enough to be evidence.
So item 3 carries the pitch, and is the one to test: does a prototype actually close a decision, or does it relocate it? A prototype behind a feature flag answers "can this be built" and leaves "should we operate this" untouched. Those are different decisions and only the first is cheap.
Item 2 is the one with teeth in practice. Approval gates are rarely information problems; they are accountability placements. An interface that automates a gate has either genuinely absorbed the accountability or has just routed around the person holding it, and the second is indistinguishable from the first until something fails. Worth a question if there is time.
Item 4 is the honest one, and its presence is a good sign: a method that names its own escape hatch is usually a method somebody has actually run.
The speaker
Will Larson writes at lethain.com ("Irrational Exuberance", running since 2007) and is one of the more durable practitioner-writers on engineering leadership. Not an academic: there are no papers. The corpus is four books, a long blog, and conference talks, and treating it as practitioner literature rather than research is the right frame for reading it.
Career: CTO at Imprint today; previously CTO at Carta and at Calm, and engineering leadership at Stripe, Uber and Digg. Computer science at Centre College, then a year in Japan on the JET Programme. Note that several published bios — including O'Reilly's own on the newest book — still say "CTO at Carta"; Imprint is current per his own site and LinkedIn.
Books
| year | title | publisher | what it is for |
|---|---|---|---|
| 2019 | An Elegant Puzzle: Systems of Engineering Management | Stripe Press | team sizing, technical debt, succession; the one people quote |
| 2021 | Staff Engineer: Leadership Beyond the Management Track | self-published | the senior IC track, built on interviews with people on it |
| 2024 | The Engineering Executive's Primer | O'Reilly | the executive layer above both of the above |
| 2025 | Crafting Engineering Strategy: How Thoughtful Decisions Solve Complex Problems | O'Reilly | a repeatable process from diagnosis to rollout |
Read Crafting Engineering Strategy first for this session. Its subtitle is "how thoughtful decisions solve complex problems", the webinar is "decisions rather than dates", and the case material is drawn from Stripe, Uber and Calm. The webinar is almost certainly that book's argument with the AI claim layered on top, so the book is the version with the reasoning intact.
There is also an AI Companion to Crafting Engineering Strategy published separately, which is itself a small data point about how the strategy argument and the AI argument are being packaged.
Talks and video
| what | where | note |
|---|---|---|
| Solving the Engineering Strategy Crisis | QCon SF 2024 | the substantive one; speaking notes published alongside the video |
| Leading with policy, consensus, and conviction | LeadDev | three leadership styles, framed as what his own toolkit was missing |
| A masterclass in engineering leadership | First Round Review | long-form on the Carta/Stripe/Uber/Calm arc |
| The engineering mindset | Lenny's Newsletter | the widest-reach interview |
| The Engineering Executive's Playbook | Breaking Changes, with Jean Yang | around the 2024 book |
If only one: the QCon SF 2024 talk. It is the closest thing to a primary source for the position this webinar is selling, it predates the AI framing, and the published speaking notes make it readable in ten minutes.
Papers
None. He is a practitioner-writer with no peer-reviewed publication record, and the books are not research reports — they are argued from experience at named companies, without the controls that word would imply. Stated explicitly here because "books, videos and papers" is a reasonable thing to ask and the honest answer to the third is zero.
Why it is on the calendar
The decisions-over-dates argument is a planning-side statement of something this site keeps arriving at from the engineering side: progress is measured in resolved uncertainties rather than elapsed intervals. Worth checking against grinding a PRD until it breaks, which is the same claim run destructively — take the plan and find where it stops surviving contact.
Notes
To be filled in after the session, or from the recording.