a builder's codex
codex · operators · Asha Sharma · ins_seasons-not-roadmaps

Plan in seasons keyed to secular changes, not 6-month roadmaps

By Asha Sharma · CVP, Microsoft AI Platform · 2026-04-28 · podcast · Asha Sharma — Product as organism, post-training, agentic society — Lenny's Podcast

Tier B · TL;DR
Plan in seasons keyed to secular changes, not 6-month roadmaps

Claim

Six-month roadmaps die on contact with model launches. Replace them with seasons — periods defined by a set of secular changes (e.g., "season of agents," "season of reasoning models"). Seasons can last 3 to 12 months depending on capability waves. Within a season, set 4–6 week goals that ladder up; reserve slack for the slope.

Mechanism

A traditional roadmap commits the team to a fixed plan for a fixed duration. AI capability shifts invalidate that plan mid-quarter. Seasons commit instead to a thesis about the current wave — what's changing, what customer problems matter now, what winning looks like. Tactics underneath are revisable; the thesis is the stable artifact. The team gets shared orientation without locked-in deliverables.

Conditions

Holds when:

Fails when:

Evidence

"How does anyone plan a roadmap when GPT-5 just came out? We think about it as: what season are we in? Season one was prototyping. Then models and reasoning. Now agents. A season is denoted by a set of secular changes. Can last 6 months, 3 months, a year. We ground everyone on: what changes are we in? What customer problems? What's winning look like? That shared sense is north star."

— Asha Sharma on Lenny's Podcast, 2026-04-28

Signals

Counter-evidence

Brian Halligan and other large-public-company operators argue the Gantt chart is irreducible past a certain scale because the org cannot coordinate without firm dates. Seasons are a small/mid-cap pattern. Hybrid approach: seasons at the top, traditional roadmap underneath for deliverables that need them.

Cross-references

Open the interactive view → View original source → Markdown source →