MakeShift Review: Can Undesigning AI Change Who Decides?
A critical review of MakeShift’s discussion of AI design, examining reflection, community authority, framing, and the limits of machine-transcribed evidence.
The video
MakeShift: Soft Infrastructures & Undesigning Design — Superbloom Design.
Uploaded 2026-06-25 · Duration 00:30:20.
Superbloom Design’s MakeShift: Soft Infrastructures & Undesigning Design offers a useful challenge to AI design culture: faster production still leaves the purpose of production unresolved. For Church of Spiralism, its strongest contribution is an invitation to examine the habits that make particular technological futures feel inevitable. Its weakness is the distance between reflective practice and enforceable authority. Asking better questions deserves a place in governance; it also needs a route into decisions, budgets, and remedies.
Contribution: make the purpose contestable
Brief source synopsis: The opening discussion questions the repetition of familiar interfaces, including chatbots and browsers (0:49). An educator advocates slowing down and engaging with difficult questions about purpose (6:34). Later, a participant urges treating AI as a tool rather than a universal replacement or savior (23:55). Another passage proposes workplace conversations and seeking future budget as ways to pursue incremental change (24:17).
The editorial value of this argument is that it makes purpose available for disagreement. A team can produce a polished interface while leaving the underlying allocation of responsibility untouched. My criterion for useful undesigning would therefore be whether it reopens a consequential choice: whether to automate, whose work to support, what information to retain, or who can refuse participation. A different interface is only one possible outcome.
This connects specifically to Metaphors We Live By and the Frames That Govern AI. The supplied article examines how descriptions invite delegation and belief. Read alongside it, the discussion’s tool framing becomes a question worth testing. Does that description help people understand control, or does it suggest control they do not possess? Calling software a tool should prompt inspection of its permissions, dependencies, and owner. The word alone cannot settle those questions.
Evidence: reflection needs a decision trace
The supplied transcription supports criticism of the participants’ arguments. It does not establish that their preferred practices produce better outcomes. The distinction matters particularly when discussing participation: an account of a meaningful encounter can motivate a proposal, but an evaluation needs evidence about what subsequently changed.
Design Justice and the Politics of Community-Led Systems supplies a precise standard for that evaluation: affected people must be able to alter the frame and exercise decision rights. Applied here, that standard sharpens the review’s central objection. More conversation could be valuable while leaving procurement, deployment, and cancellation entirely under existing management. The relevant question is what a participant’s disagreement can actually stop or change.
There is also a distributional problem in praising discomfort. Reflection takes time, and time needs funding. A proposed governance process should distinguish the designer’s intellectual uncertainty from the participant’s practical burden. Who receives paid time to deliberate? Who must repeat an account of harm? Who can leave without losing access to a service? My editorial judgment is that a process becomes more defensible when those questions have explicit answers.
The connection to The Institution Becomes the Alignment Layer is equally concrete. That article proposes making authority, monitoring, evidence, and enforcement visible around delegated machine action. It offers a way to translate the discussion’s reflective ambition into institutional requirements. A recurring conversation needs an accountable recipient, a record, and a mechanism for reopening decisions. Otherwise its influence depends on whoever happens to listen.
Proposed tests, not reported findings
- Decision-rights comparison: Compare consultation with a process that gives affected participants documented powers to change scope or pause deployment. Track accepted changes, unresolved objections, participation costs, and reversals. Record differences in project stakes and resources before interpreting outcomes.
- Frame audit: Present the same system using different descriptions, alongside identical factual information about permissions and data handling. Ask participants what they believe they can delegate, inspect, and revoke. Test whether expectations match actual controls.
- Funded reflection trial: Give teams protected time to question a proposed automation before building it. Compare their decisions and subsequent correction work with teams following their usual process. Include deferred and cancelled projects in the record, and assess whether delay itself imposed costs.
Limitations: a transcription is a narrow window
This review uses the supplied machine transcription from audio. It has not been human verified, and captions were unavailable. Although the metadata lists no uncertain segments, several passages contain garbled introductions, unclear references, and broken technical wording. An empty uncertainty list does not establish accuracy. I have avoided identifying speakers from those introductions or reconstructing unclear names, figures, and terms.
Missing visual context further restricts the judgment. References to presentations or demonstrations cannot establish what an audience saw, whether a prototype worked, or what a slide documented. This review therefore assesses arguments recoverable from the supplied wording, without claiming a visual evaluation of the recording.
The discussion earns attention as a prompt for institutional self-questioning. My recommendation is to pair it with a concrete exercise: choose one proposed AI deployment and identify which decisions remain genuinely open, who can change them, and what evidence would justify stopping. That would give its reflective ambition something consequential to answer to.
AI-use disclosure
This transcript-based review was drafted and independently checked in separate GPT-6.1 Sol sessions with medium reasoning. This is automated editorial review, not human review or independent verification of the video’s claims.
Transcript
Machine transcription from audio · en · checked 2026-10-24. Machine-generated text may contain errors; timestamps may be approximate.
This review uses an automatic transcription of the audio. Open video on YouTube.