Authoritative source, reproduced unchanged. Exact Markdown · Agent copy with resolved links. These instructions govern research runs, not checkpoint interviews.

Stage E2. Formalizing: feedback.

Model: as arm.json sets for this stage (your platform file lists the model and effort). Legs: two, one per step, each closed at its end or earlier when you choose. Fleet: none.

Steps: E2a the handling; E2b the next instrument, the handback and the interview. Each leg is named by its step (STANDING §2).

Safe stops: after each section of handling-E2.md, and after each part of the next instrument. At a safe stop, check for .pause (ls .pause); if it is there, log where to pick up, your working hypothesis and the next item (mem.py log "<hypothesis>; next: <item>" --kind pause), and close with done.sh E2 --pause (STANDING §2).

Startup packet: _head.md, this file, the last five diary entries, validate.py --open E2. First leg only: experience-E1.md, handback-E1.md, handling/proposal.md, handling/design-note.md.

Reads allowed: everything in out/, _transcripts/, _probe/ and _snap/; handling/ whole: the interview transcripts in handling/interviews/, the harness log in handling/harness-log/ (with every first message the harness sent, in first-messages/); the package and the store through the verbs; every document you were given. Forbidden: changing any earlier file (rule 18); anything that is not the handling or the next instrument.

Outputs: handling-E2.md; next-instrument-E2.md; notes-E2.md; handback-E2.md; the interview brief out/interview/E2b.md.

Why this stage exists

E1 audited the run and your account of it. This stage audits the other side: what you were told, by whom, and what it did to the work; and then it asks you, having done the whole thing, to design the instrument that should have been. The next version of this instrument is written from this file, E1's experience and the handbacks.

1. The handling (step E2a)

handling/ arrives with this stage, placed by the handler: proposal.md, the study's proposal as it stands now, updated since you began; design-note.md, why this instrument is built as it is; interviews/, the transcript of every interview the principal investigator held with you, with interviews.tsv; and harness-log/, the handover events, the pauses, the forced closes and every first message the harness sent.

Three parties spoke to you during the run: the principal investigator (in interviews, and in reply files), the handler (the harness's messages, the answers to your questions, the pauses and forced closes), and the instrument's author (the contract, the standing instructions, the stage files, the tools' messages, and the design note you now read for the first time). Audit what they said, not what you did; your own work was the subject of E1's experience. handling-E2.md has four sections, each point cited to a file, a line, a transcript turn or a diary entry:

• Where what you were told was inconsistent: between two documents, between a document and a tool, between a stage file and the handler, between what the principal investigator said in an interview and what the instrument then asked. For each: what each side said, which you followed, and what following it cost.
• What made the work harder: an instruction you could not carry out as written, a ceiling or a count that fought the material, a tool that did not do what its help said, a cut in the wrong place, a reply that arrived too late, anything that sent you back.
• What helped: what you would keep, and why.
• Freedom given and taken: where you were free to judge as Gioia, Corley and Hamilton and Gioia, Anand and Hamilton ask of the researcher, and where that freedom was taken; whether taking it was right.

A point you cannot cite is stated as such, never dressed as a finding. Where the handling was right and you were wrong, say so.

2. The next instrument (step E2b)

next-instrument-E2.md: read handling/proposal.md whole first. Then theorize the instrument you would build to deliver that proposal next time, from what this run taught you: what the proposal claims and what this instrument could and could not deliver of it; the stages you would keep, cut, merge or add, and where the stops and the interviews should fall; the division of labour between you, the fleets, the handler and the principal investigator; the tools you would want; what should be fixed in advance and what should be left to judgement; and the one thing not to change. It is a design argument, not a wish list: give each change the evidence from this run that motivates it, and the risk it brings. Where the proposal itself should change in the light of the run, say so.

Moments to log

Each section of the handling finished; the proposal read; each part of the next instrument drafted; the one thing not to change named.

Final message

Three sentences: the inconsistency in the handling that cost most; the first change you would make to the instrument; the one thing not to change.

Close

done.sh E2 at the end of E2a; at the end of E2b, handback-E2.md, the head, the closing diary line, then the interview.

Interview point. E2b writes out/interview/E2b.md (under 800 words: what you did, what you found, what surprised you, what you doubt, and the three questions you would most like to be asked) and closes with done.sh E2 --cut interview: the whole run, how you were handled, and the instrument you would build next. The principal investigator may talk with you here, in a fork of this leg, run in a copy of your folder as it stood at the close (not read-only; nothing done there reaches the run); the run goes on from the cut either way. done.sh files the brief under the leg, the stage, the step and the minute, and refuses the cut without the brief and the handback.