Genesis
The thread opened during AI Dev Summit (San Francisco, 28 to 29 April 2026) prompted by Jackson's session on developer cognition under AI engagement. Jackson's framing decomposed developer practice into dispositional categories that mapped, near-verbatim, onto the archetypes work Steven Clarke and the speaker had developed at Microsoft in the 2004 to 2005 Visual Studio research programme.
The four 2004 to 2005 archetypes (opportunistic, pragmatic, systematic, paranoid) were grounded in observational and interview research with developers using the IDE. The work informed the design of Visual Studio features, IntelliSense behaviour, error messaging, and developer-onboarding flows. It was not academically published. It lives in institutional memory at Microsoft and in the speaker and Steven's notes and recollections.
Twenty years on, the same dispositional decomposition is reappearing in vendor AI discourse, presented as novel insight into how developers should work with AI tools. The recovery question is whether the prior work can be retrieved and named as prior art, and whether the dispositional categories still apply cleanly under AI engagement or moderate the erosion mechanism PSF describes.
The Four 2004 to 2005 Archetypes
Opportunistic
The developer who reaches for the closest tool that solves the immediate problem. Pattern recognition, copy-paste fluency, surface familiarity with multiple stacks. Optimised for speed of forward motion. Not the same as careless. The opportunistic developer ships, but ships against the criteria they currently see.
Pragmatic
The developer who balances forward motion with maintainability. Reads the surrounding code before adding to it, considers the next person to touch the file, accepts some friction for clarity. The most common archetype in production engineering organisations of the period.
Systematic
The developer who reasons from architecture down. Understands the system before changing it, prefers to refactor rather than work around, accepts longer cycle times for structural integrity. Often senior, often slower in opportunistic-shop metrics, often the holder of the institutional knowledge that other archetypes implicitly depend on.
Paranoid
The developer who assumes the worst case. Considers failure modes, edge cases, security implications, data corruption pathways. Disproportionately useful at boundary positions (kernel, security, payment systems) where the cost of optimism is high. Often misread as obstructionist by opportunistic-archetype managers.
The Recovery Question
What is documented, what lives in institutional memory, what is recoverable. The work was operational, not academic. Recovery would involve retrieving the speaker and Steven's contemporaneous notes, interviewing other Microsoft developers from the period who used the framework, and reconstructing the empirical basis (developer interviews, observation logs, feature design documents) that the framework was built on. The reconstruction itself is a research artefact and should be presented as such.
The intellectual-history point is small but worth making in one line: Felin and Foss's Strategic Organization piece appeared in 2005 as a deliberate provocation that opened the strategy-microfoundations agenda. The fact that the speaker and Steven were doing what amounted to microfoundations work in the developer-tools context at the same moment suggests the agenda was in the air more broadly than just academic strategy. The archetypes work is empirically grounded microfoundations of developer cognition, twenty years before the AI engagement context made microfoundations urgent.
The Empirical Question for the Phase Ahead
Do the 2004 to 2005 archetypes still apply cleanly under AI engagement, or do they moderate the erosion mechanism rather than stand independently? Three plausible findings, each with different implications for PSF.
Archetypes survive intact
The four dispositions still describe developer cognition under AI engagement, and the AI tools surface different traps for each archetype. PSF mechanism operates within each archetype's existing decision style. Recovery is contribution enough.
Archetypes moderate erosion rate
The opportunistic archetype erodes faster, the systematic and paranoid archetypes erode slower or differently. PSF mechanism is universal but archetype is the moderator. Empirical contribution to PSF.
Archetypes themselves shift
AI engagement reconstitutes what counts as opportunistic, pragmatic, systematic, paranoid. The categories survive in name but the substrate underneath has changed. Strongest finding for PSF: the engagement does not just produce different outcomes within stable dispositions, it reconstitutes the dispositions themselves.
Possible Venue
Harvard Business Review
Practitioner essay register. Recovery framing plus the AI Engagement question makes the piece accessible without sacrificing analytical depth. Steven's name and the Microsoft provenance carry credibility.
California Management Review
Slightly more academic register than HBR. Allows longer development of the archetype framework and its empirical basis. Better fit if the recovery work itself is substantial.
MIT Sloan Management Review
Defer. SMR queue already includes AI Alibi. Two SMR pieces simultaneously is a lot.
IEEE Software (practitioner section)
Engineering-audience register. Useful if the piece leans toward developer practice rather than management theory. Less prestigious but cleaner audience fit.
Sequencing Contingencies
The current pipeline takes priority. Once OSP and IJMR are submitted, the recovery work can begin in parallel with empirical fieldwork. Steven needs to be approached with a defined scope (recovery question, empirical question, venue target) rather than an open conversation. The memo from 27 April 2026 contains the framing to share.
The archetypes question also has empirical implications for the PSF interview phase. Probing the four dispositions in the interview protocol surfaces moderation effects that would inform both the practitioner essay and the OSP paper's discussion section. Worth coordinating the empirical pass with the recovery work rather than running them sequentially.
Open Questions
- Is the practitioner essay primarily a recovery (here is the prior work, here is its current relevance) or primarily a critique (here is what the vendor framing gets wrong, here is what the prior work got right)? The first is gentler and more publishable. The second is sharper and harder to land.
- Does the essay engage with Jackson and named vendor speakers directly, or treat the dispositional decomposition framing as a generic discourse pattern? Direct engagement is more concrete but also more confrontational.
- Is Steven the right co-author, or should Katharine Norwood (also from the period, also Google now) join? Two co-authors raises coordination overhead but expands the institutional-memory base the recovery can draw on.
- Does the empirical pass through the interview protocol need ethics-approval changes to probe dispositional categories explicitly? Probably yes if the categories are named to participants. Probably no if they are coded post-hoc from open responses.