Add burn-down between rounds as an SDI technique #23

New issue

Closed

opened 2026-05-23 20:55:41 +00:00 by coilysiren · 1 comment

coilysiren commented

2026-05-23 20:55:41 +00:00

Owner

Originally filed by @coilysiren on 2026-04-29T04:50:20Z - https://github.com/coilysiren/website/issues/1329

🤖 Filed by Claude Code on Kai's behalf.

The Stochastic Design Iteration post describes generate-variance and human-filter as the two main moves in a round. A third move surfaced in a recent session and is worth adding: burn-down between rounds.

Burn-down is the act of compressing the WIP doc between iteration rounds. Strip scaffolding (status banners, conversation-orientation prose, sprawling TODO lists), leave only load-bearing substrate. The next round then perturbs a clean dense file rather than rewording fluff.

Why it matters:

Without burn-down, prose padding accumulates and each round wastes variance on rewording filler.
Burn-down forces the human to re-confirm what is load-bearing. Anything that survives a burn-down has earned its place.
The compressed file is closer to "ready for fresh-context iteration," which is the SDI ideal.

Open questions for the post to resolve:

Who initiates burn-down. In the session that surfaced this, the human called for it and the LLM executed. Plausible alternative: the LLM initiates burn-down whenever it notices the WIP has drifted into prose-heavy territory.
Cadence. After every N rounds, or whenever the human notices fluff, or both.
Whether burn-down is one move or two (compress + restructure).

Companion: see #1327 on SDI/ALP as a coherent collection. Burn-down is a candidate technique to surface in either the SDI post or a sibling post on SDI techniques.

🤖 Filed by Claude Code on Kai's behalf.

_Originally filed by @coilysiren on 2026-04-29T04:50:20Z - [https://github.com/coilysiren/website/issues/1329](https://github.com/coilysiren/website/issues/1329)_ > 🤖 Filed by Claude Code on Kai's behalf. The [Stochastic Design Iteration post](https://coilysiren.me/posts/stochastic-design-iteration) describes generate-variance and human-filter as the two main moves in a round. A third move surfaced in a recent session and is worth adding: **burn-down between rounds.** Burn-down is the act of compressing the WIP doc between iteration rounds. Strip scaffolding (status banners, conversation-orientation prose, sprawling TODO lists), leave only load-bearing substrate. The next round then perturbs a clean dense file rather than rewording fluff. Why it matters: - Without burn-down, prose padding accumulates and each round wastes variance on rewording filler. - Burn-down forces the human to re-confirm what is load-bearing. Anything that survives a burn-down has earned its place. - The compressed file is closer to "ready for fresh-context iteration," which is the SDI ideal. Open questions for the post to resolve: - Who initiates burn-down. In the session that surfaced this, the human called for it and the LLM executed. Plausible alternative: the LLM initiates burn-down whenever it notices the WIP has drifted into prose-heavy territory. - Cadence. After every N rounds, or whenever the human notices fluff, or both. - Whether burn-down is one move or two (compress + restructure). Companion: see [#1327](https://github.com/coilysiren/website/issues/1327) on SDI/ALP as a coherent collection. Burn-down is a candidate technique to surface in either the SDI post or a sibling post on SDI techniques. > 🤖 Filed by Claude Code on Kai's behalf.

coilysiren commented

2026-05-30 05:42:59 +00:00

Author

Owner

Iceboxed in the 2026-05-29 backlog burn-down: Add a technique to a blog post. Reopen anytime if it becomes real.

coilysiren

2026-05-30 05:43:00 +00:00