ABSORB v0.2.0 (draft)

ABSORB v0.2.0 makes the assessment a working tool for a team, not just a questionnaire. The statements, scoring, level rule and tier thresholds are unchanged, so v0.1.x results compare directly.

New in the assessment page

• Pipeline view. Each stage of a typical change is shown as a block, split into hands-on work and waiting. Sliders show what happens to the whole path when one stage's work gets faster or its waiting is cut. On the example data, making Build 10× faster gives 1.04× end to end; halving the waiting at the constraint gives 1.14×.
• Compare with a colleague. Each result has a short code. Paste a colleague's to see both raters side by side, the average gap against the H5 aim of 0.5, and the statements where you differ by 2 or more points under "Talk about these first".
• Decision record. Record the constraint you believe in, one change, and a prediction with a date. Save a snapshot; at the next assessment, see whether the prediction was met and whether the constraint moved (H4).
• Guided mode. One statement at a time, keyboard-friendly, with an evidence hint for each dimension.

Everything still runs offline in a single HTML file. Nothing is sent anywhere.

Spec and validation

• The speedup formula gains an optional waiting term, for exploring only.
• VALIDATION.md explains how the page collects data for H4 and H5.

See CHANGELOG.md for details. Cite all versions with the concept DOI 10.5281/zenodo.23895787.

ABSORB v0.1.1 (draft)

Bug-fix release for the assessment page. No change to the framework, scoring or wording, so v0.1.0 results remain valid.

• Tabs fixed. The Answer panel always showed below Results, because a display rule overrode the hidden attribute. Results and Answer now switch as intended.
• Safari text fixed. Added <!DOCTYPE html> and a UTF-8 character encoding. Without them Safari showed "×" and "·" as garbled characters, and browsers rendered the page in quirks mode.
• Removed a leftover meta tag from the drafting tool.

Tested in Chromium and WebKit: no errors, no network requests.

Cite all versions with the concept DOI 10.5281/zenodo.23895787.

ABSORB v0.1.0 (draft)

First public draft of ABSORB (Agentic Bottleneck, Speed & Output Readiness Baseline): a diagnostic for whether an organization can turn faster AI-driven coding into faster, safer delivery.

Status: draft v0.1. A proposed framework, not a validated standard. Levels, tiers and thresholds are proposals. VALIDATION.md states in advance what would count as evidence against them.

What's included

• SPEC.md covers the method, eight dimensions, maturity levels 0–4 and autonomy tiers T0–T3. Its sources table records how each figure was checked.
• assessment/index.html is an offline self-assessment with 24 statements, stage timing split into hands-on time and waiting, a whole-path speedup check, and tier readiness. It makes no network requests.
• VALIDATION.md lists eight hypotheses with falsification criteria and four study designs.
• PRIOR_ART.md explains how ABSORB relates to DORA 2025 and other readiness frameworks. The "AI as amplifier" thesis is DORA's. ABSORB adds the stage-timing speedup check, the evidence-based autonomy tiers and a falsifiable instrument.
• case-studies/TEMPLATE.md is a template for documenting an application. Critique and case studies are invited.

See CHANGELOG.md for details. Drafted with AI assistance (Claude).