Research Journal Entry

Objective

Test whether current evidence supports a universal productivity claim for AI coding assistance.

Starting state

All 39 repository Markdown files were read. No accepted domain evidence, hypothesis, theory, experiment, or REP existed. Baseline registry and validation checks passed.

Actions taken

Selected a bounded first topic; searched recent primary research; compared a vendor controlled experiment, independent realistic RCT, organizational study, holistic evaluation, and methodological update; created claim-level records.

Observations

Effect estimates reverse across settings. Perception can diverge from measured time. Individual and delivery-level outcomes can diverge. Automated correctness does not guarantee merge readiness. Ubiquitous adoption makes abstention experiments harder.

Evidence collected

EV-AICODING-2026-10A1, EV-AICODING-2026-3B8F, EV-AICODING-2026-6E21, EV-AICODING-2026-C940, and EV-AICODING-2026-F5D2.

Hypotheses considered

Universal positive effect; universal negative effect; time-only explanation; conditional net-value effect.

Attempts to falsify

Used the strongest positive controlled result against a universal-negative claim and the strongest negative realistic result against a universal-positive claim. Neither universal claim survived its boundary conditions.

Decisions and rationale

Adopted HY-AICODING-2026-7C2D as supported and TH-AICODING-2026-44D0 as a candidate, both at Medium confidence. No practitioner deployment mandate was issued.

Failures and dead ends

This was not a systematic review. Full-text extraction was incomplete for the DORA report. Studies are not sufficiently harmonized for meta-analysis.

Confidence changes

From no justified belief to Medium confidence that effects are heterogeneous; Low-to-Medium confidence in any named moderator's causal contribution.

Files changed

Mission, five evidence records, hypothesis, theory, proposed experiment, journal, REP, current context, queue, and handoff.

Highest-value next step

Review the REP, preregister EX-AICODING-2026-2D77 with a participating organization, then run a telemetry pilot.