positioning · working document

What You Study, in Words Other People Already Use

A thesis, the real fields it maps to, three lengths of answer for when someone asks, and an honest plan — including what to stop doing.

The thesis

I study what machine mediation does to human judgment — and I build interactive instruments that let people feel those mechanisms instead of being told about them.

Two halves, and both are load-bearing. The first half is the subject and it is crowded — everyone has takes about attention and AI. The second half is the differentiator, and almost nobody has it: the ability to build a thing where you drag a slider and watch a population’s mean slide out from under you. Lead with the subject; get hired for the method.

The fields, honestly named

field what it is · who to cite · why it’s yours
postphenomenology The core one. How technologies mediate the human–world relation — not what tools do for us but what they do to the relation itself. Don Ihde, Peter-Paul Verbeek. This is the actual academic name for the thing you kept calling “phenomenology.” Use it and you sound situated instead of searching.
4E cognition Embodied, embedded, enacted, extended mind. Clark & Chalmers, Andy Clark’s predictive-processing work. The empirical wing of your delegation question — what actually happens to a mind that offloads. You’re already reading here; claim it.
explorable
explanations		 Your method, and it has a name. Bret Victor’s 2011 essay, Nicky Case, Distill.pub, The Pudding, Our World in Data. Interactive artifacts that teach a mechanism by letting you operate it. Your Galton board with the bias slider is textbook. This is the most under-claimed thing about you.
media ecology McLuhan, Postman: environments-not-messages. Older, less fashionable, and the direct ancestor of what you mean by “the world changed and our instruments didn’t.”
political economy
of attention		 Zuboff, Crawford, Wu, Whittaker. The structural half — who profits from the mediation. Use this instead of “intersectionality of capitalist structures,” which means something else and will get you corrected.
sociotechnical
AI safety		 The non-technical wing of safety: societal impact, epistemic security, information integrity. Distinct from alignment research, and the distinction matters — see the right column.

What you can claim, and what you can’t (yet)

defensible today

  • Builder — shipped systems, protocols, evals, MCP tooling. Receipts exist.
  • Explorable-explanation author — the figures are real work in a real genre.
  • Essayist on machine mediation — once the essays are actually public.
  • Self-experimenter / n=1 methodologist — the biometric and credence work is genuinely unusual.
  • Well-read amateur in philosophy of mind and phil-of-tech. “Amateur” is not an insult in this lineage; it’s accurate and disarming.

not yet — don’t say it

  • “AI safety researcher.” You did fundamentals, not research. Say “studied AI safety fundamentals” until there’s output.
  • “Philosopher.” No training, no publication. Say you work in philosophy of technology.
  • “Alignment.” Your interest is societal and cognitive, not technical alignment. Claiming it invites one question you can’t answer.
  • Invented terms — “noegraphy” reads as a tell. Coin words only after you’re known, never before.

Say it out loud — three lengths

at a party · 8 words“I build toys that show what tech does to thinking.”
bio line · 25 wordsParam Vaswani builds interactive instruments for thinking clearly under machine mediation — small tools, essays, and self-experiments on what we lose when judgment gets outsourced.
application · 60 wordsI work at the intersection of philosophy of technology and interactive explanation. My subject is what constant machine mediation does to human judgment — delegation, calibration, and the political economy of attention. My method is building: small instruments that let someone operate a mechanism rather than read a claim about it, plus essays and n=1 experiments that hold themselves to published credences.

The manifesto

six lines · say these in public
  1. Checkability, not certainty.Cheap intelligence didn’t make the world knowable. It made it verifiable. The goal was never to be sure — it was to never be gaslit.
  2. The environment changed; the instruments didn’t.We are running attention hardware built for scarcity inside an economy engineered for abundance. Index for that.
  3. Show the mechanism, don’t assert the conclusion.A claim you can operate beats a claim you must accept. If it can’t be built as a toy, it probably isn’t understood.
  4. Self-knowledge that changes nothing is decoration.Insight has to reach the calendar, the diet, the decision. Otherwise it’s a personality quiz.
  5. Anything you can’t leave, owns you.Exit is a design requirement, not a mood. This applies to platforms, jobs, frameworks, and beliefs.
  6. Hold the contradiction; name it when it won’t resolve.Both/and is not indecision. Forcing resolution destroys more information than it creates.

What to actually do

the honest part

Your constraint is not clarity. It is published work. You have spent roughly a month on positioning — a masthead, a moniker, a surfaces plan, nine versions of one homepage — and today you asked for a manifesto. Meanwhile your public surface is approximately zero: Substack deleted, GitHub deleted, domain unregistered, essays unpublished.

The thesis above is good enough. It will not get better by being refined again; it will get better by being tested against readers. Another positioning document is the most comfortable possible way to avoid that test, and you should treat this one as the last.

  1. Register the domain.today · 5 minNothing downstream works without it, and it has been “register-ASAP” since early August. It’s not a strategy decision, it’s a checkbox blocking six other things.
  2. Ship three explorables in ninety days.by ~1 decNot essays — explorables, because that’s your edge and nobody else in your lane has it. You already have the engines: (a) The Delegation Discount — the Galton bias machine, generalized to judgment atrophy; (b) The Sameness Engine — a recommender sim where you watch a population’s preferences converge on themselves; (c) Calibration Gym — make predictions, get Brier-scored, see your overconfidence by domain. Each is a weekend or two. Each doubles as a portfolio piece and an essay.
  3. Publish where the readers already are.with each dropPost to LessWrong and Hacker News, not just your own site. An explorable with real math is one of the few things that reliably travels in those rooms — and those rooms contain the people who hire for the work you want.
  4. Stop redesigning the site.nowIt is done. It was done two versions ago. Ship the three fixes in the handoff, deploy, and freeze it until there’s new work to put on it. The site is a container; you are short on contents, not container.
  5. Re-decide the credentials, honestly.before dec 15Columbia J-School at ~$131K unfunded is a poor trade for someone whose differentiator is building interactive artifacts — journalism school teaches reporting, not this, and the industry it credentials you for is contracting. A philosophy PhD is a five-year detour from work you can do now. If you want a credential, the cheap one is three published explorables and a talk. Keep BlueDot; it’s short, relevant, and the deadline is imminent.

How you’ll know it worked

Not likes. Three checkable things, on a date: one explorable that a stranger you don’t know reshares; one inbound conversation you didn’t initiate; and one instance of someone else using your framing back at you in their own words. Any one of those is worth more than a fourth positioning pass. If none of the three happens by March, the thesis is wrong or the distribution is wrong — and you’ll be able to tell which, which is the whole point of writing it down now.