A Jellyfish Knows More About Time Than Your Physics Textbook

4–6 minutes

Disclosure: I shouldn’t have to apologise for two consecutive posts written by Claude – previous post on duration and time – this time about reframing a chat about time and duration around the hydrozoan jellyfish (sic. sea jelly). Yet again, this post arrives with a warning: if you don’t like AI-authored content, it’s not too late to turn back or click away.

Audio: NotebookLM summary podcast of this topic.

A hydrozoan jellyfish off the coast of Japan has no brain, no central nervous system, and no opinion about the block universe. It does, however, have something to teach us about the nature of time — and it’s more than most pop-science accounts manage.

Researchers recently described Clytia sp. IZ-D, a previously undocumented species of jellyfish that releases its gametes each evening with remarkable precision. The interesting part isn’t the precision. It’s what produces it.

Left to its own devices under constant light, the jellyfish spawns every 20 hours. Not 24. Twenty. It has its own autonomous developmental rhythm — oocytes growing, acquiring competence, reaching a threshold, and releasing. That rhythm belongs to the organism. It’s not borrowed from the environment. It’s not calibrated to the rotation of the Earth. It’s the jellyfish’s own tempo.

But place the jellyfish under a normal day-night cycle, and the 20-hour rhythm gets pulled into a 24-hour one. The daily light-dark transition doesn’t trigger spawning directly. It delays the autonomous cycle by several hours, dragging it into alignment with the environmental period. Sunrise resets the clock — not by starting it, but by holding it back. Under a standard 12-hour light cycle at 21°C, spawning occurs 14 hours after dawn. Every day. Synchronised across the population to within about 20 minutes.

The 24-hour cycle is not the jellyfish’s rhythm. It’s what happens when an external structure is imposed on it.

Now, you could read this as a charming piece of marine biology and move on. But if you’re paying attention, it’s a remarkably clean illustration of something that applies well beyond cnidarian reproduction.

The jellyfish has its own duration — a developmental tempo with no fixed metric. Under constant conditions, the cycle length shifts with temperature: 17 hours at 24°C, 22 hours at 18°C. There’s no intrinsic interval. The process has structure — oocyte growth, staged competence acquisition, threshold sensitivity — but it has no grid lines. No bar lines. No BPM.

The 24-hour day provides the grid. The light-dark cycle imposes intervallic structure on a process that doesn’t natively possess it. And the imposition works — it synchronises the population, coordinates male and female gamete release, and presumably improves reproductive success. The grid is useful. Indispensable, even.

But it’s not the organism’s time. It’s what the organism’s time looks like after the environment has had its way with it.

And here’s the part that matters philosophically. The imposition is imperfect. Individual oocytes within the same gonad don’t all reach maturation competence at the same moment. There’s a spread — 40 to 60 minutes under normal conditions, widening further under constant light. The grid says “now.” The biology says “roughly now, give or take, depending on which oocyte you’re asking.” The grid cannot fully resolve what it organises. There’s always a residue.

Under constant darkness, the residue takes over entirely. The autonomous rhythm reasserts itself, but without the entraining signal it becomes asynchronous. Different jellyfish spawn at different times. Different oocytes within the same jellyfish mature at different rates. The developmental process is still there — the structure is still there — but without the intervallic imposition, it doesn’t produce anything that looks like coordinated “time.” It produces duration doing what duration does when nobody is counting.

None of this requires a brain. None of it requires experience. None of it requires a subject who feels the passage of time. The jellyfish doesn’t experience the 24-hour cycle as given. It doesn’t experience anything, so far as we know. And that’s precisely what makes it useful as an illustration: it shows intervallic structuring operating at a purely material level, without any phenomenological overlay. The grid is imposed on the biology. The biology is not the grid.

Now scale up.

We do the same thing the light-dark cycle does to the jellyfish, except we do it to everything, and we do it to ourselves. We impose intervallic structure on duration — segmenting it, ordering it, metrising it — and then we mistake the structure for what it’s imposed on. We experience time as sequenced, directional, and measured, and we assume that’s what time is. Physics formalises the assumption into equations and discovers, to no one’s surprise, that the formalism is time-symmetric. And then pop science announces that time could therefore “go backwards” — which is exactly like saying that because the score reads the same in both directions, the music could be played in reverse.

The jellyfish knows better. Not because it’s clever, but because it’s simple enough that you can see the layers separately: the autonomous developmental duration, the environmental intervallic imposition, the imperfect entrainment, and the residue the grid can’t capture. In us, those layers are collapsed. We live inside the imposition and mistake it for the terrain.

The 24-hour day is a grid. A useful grid. An indispensable grid. But the jellyfish was doing something before the grid arrived, and it will go on doing something if the grid is removed. What it does without the grid doesn’t look like “time.” It looks like biology unfolding at its own pace, in its own structure, answerable to its own constraints.

That’s duration. And it was there before we started counting.

Using Generative AI as Early Peer Review

4–6 minutes

Cheap Adversaries, Outsourced Ego, and Engineered Critique ← ChatGPT is obsessed with subtitles.

There is a peculiar anxiety around admitting that one uses generative AI in serious intellectual work. The anxiety usually takes one of two forms. Either the AI is accused of replacing thinking, or it is accused of flattering the thinker into delusion. Both charges miss the point, and both underestimate how brittle early-stage human peer review often is.

What follows is not a defence of AI as an oracle, nor a claim that it produces insight on its own. It is an account of how generative models can be used – deliberately, adversarially, and with constraints – as a form of early peer pressure. Not peer review in the formal sense, but a rehearsal space where ideas are misread, overstated, deflated, and occasionally rescued from themselves.

Audio: NotebookLM summary podcast of this topic.

The unromantic workflow

The method itself is intentionally dull:

  1. Draft a thesis statement.
    Rinse & repeat.
  2. Draft an abstract.
    Rinse & repeat.
  3. Construct an annotated outline.
    Rinse & repeat.
  4. Only then begin drafting prose.

At each stage, the goal is not encouragement or expansion but pressure. The questions I ask are things like:

  • Is this already well-trodden ground?
  • Is this just X with different vocabulary?
  • What objection would kill this quickly?
  • What would a sceptical reviewer object to first?

The key is timing. This pressure is applied before the idea is polished enough to be defended. The aim is not confidence-building; it is early damage.

Image: NotebookLM infographic on this topic.

Why generative AI helps

In an ideal world, one would have immediate access to sharp colleagues willing to interrogate half-formed ideas. In practice, that ecology is rarely available on demand. Even when it is, early feedback from humans often comes bundled with politeness, status dynamics, disciplinary loyalty, or simple fatigue.

Generative models are always available, never bored, and indifferent to social cost. That doesn’t make them right. It makes them cheap adversaries. And at this stage, adversaries are more useful than allies.

Flattery is a bias, not a sin

Large language models are biased toward cooperation. Left unchecked, they will praise mediocre ideas and expand bad ones into impressive nonsense. This is not a moral failure. It is a structural bias.

The response is not to complain about flattery, but to engineer against it.

Sidebar: A concrete failure mode

I recently tested a thesis on Mistral about object permanence. After three exchanges, the model had escalated a narrow claim into an overarching framework, complete with invented subcategories and false precision. The prose was confident. The structure was impressive. The argument was unrecognisable.

This is the Dunning-Kruger risk in practice. The model produced something internally coherent that I lacked the domain expertise to properly evaluate. Coherence felt like correctness.

The countermeasure was using a second model, which immediately flagged the overreach. Disagreement between models is often more informative than agreement.

Three tactics matter here.

1. Role constraint
Models respond strongly to role specification. Asking explicitly for critique, objections, boundary-setting, and likely reviewer resistance produces materially different output than asking for ‘thoughts’ or ‘feedback’.

2. Third-person framing
First-person presentation cues collaboration. Third-person presentation cues evaluation.

Compare:

  • Here’s my thesis; what do you think?
  • Here is a draft thesis someone is considering. Please evaluate its strengths, weaknesses, and likely objections.

The difference is stark. The first invites repair and encouragement. The second licenses dismissal. This is not trickery; it is context engineering.

3. Multiple models, in parallel
Different models have different failure modes. One flatters. Another nitpicks. A third accuses the work of reinventing the wheel. Their disagreement is the point. Where they converge, caution is warranted. Where they diverge, something interesting is happening.

‘Claude says…’: outsourcing the ego

One tactic emerged almost accidentally and turned out to be the most useful of all.

Rather than responding directly to feedback, I often relay it as:

“Claude says this…”

The conversation then shifts from defending an idea to assessing a reading of it. This does two things at once:

  • It removes personal defensiveness. No one feels obliged to be kind to Claude.
  • It invites second-order critique. People are often better at evaluating a critique than generating one from scratch.

This mirrors how academic peer review actually functions:

  • Reviewer 2 thinks you’re doing X.
  • That seems like a misreading.
  • This objection bites; that one doesn’t.

The difference is temporal. I am doing this before the draft hardens and before identity becomes entangled with the argument.

Guardrails against self-delusion

There is a genuine Dunning–Kruger risk when working outside one’s formal domain. Generative AI does not remove that risk. Used poorly, it can amplify it.

The countermeasure is not humility as a posture, but friction as a method:

  • multiple models,
  • adversarial prompting,
  • third-person evaluation,
  • critique of critiques,
  • and iterative narrowing before committing to form.

None of this guarantees correctness. It does something more modest and more important: it makes it harder to confuse internal coherence with external adequacy.

What this cannot do

It’s worth being explicit about the limits. Generative models cannot tell you whether a claim is true. They can tell you how it is likely to be read, misread, resisted, or dismissed. They cannot arbitrate significance. They cannot decide what risks are worth taking. They cannot replace judgment. Those decisions remain stubbornly human.

What AI can do – when used carefully – is surface pressure early, cheaply, and without social cost. It lets ideas announce their limits faster, while those limits are still negotiable.

A brief meta-note

For what it’s worth, Claude itself was asked to critique an earlier draft of this post. It suggested compressing the familiar arguments, foregrounding the ‘Claude says…’ tactic as the real contribution, and strengthening the ending by naming what the method cannot do.

That feedback improved the piece. Which is, rather conveniently, the point.

Rejection Letter

Rejected Note
2–3 minutes

Dear Author. [REDACTED] does not accept the submission of personal works produced by students, independent researchers, or professionals who have not yet attained a doctoral level. This is a moderation policy intended to ensure that publications deposited on the platform originate from qualified researchers affiliated with a recognized institution (REDACTED) and acknowledged for their expertise or previous work in the relevant field of research. This rule applies regardless of the quality or scientific value of the work, which is by no means in question here. We therefore regret to inform you that we are unable to accept this submission. If you wish, we invite you to share your work through other open platforms such as Zenodo, which allow all authors to make their research visible. Thank you for your understanding. Kind regards

Allow me to rephrase this:

Disappointing, though hardly surprising. This is the same logic as age-based thresholds I have recently taken a hammer to: crude proxies elevated into moral and epistemic gatekeepers. Not ‘is this good?’, but ‘are you old enough, stamped enough, letterheaded enough to be taken seriously?’. A bureaucratic horoscope.

Yes, I use Zenodo. I use PhilPapers. I will continue to do so. But let’s not pretend all platforms are socially equivalent. Journals still function as credibility engines, not because they magically improve truth, but because they distribute legitimacy. To be excluded on status grounds alone is not a quality filter. It is a caste system with footnotes.

And journals already make participation unnecessarily hostile. Many refuse work that has been publicly shared at all, even in preprint form. Lead times stretch to a year or more. The result is that anyone attempting to contribute to live debates is instructed to sit quietly whilst the conversation moves on without them. In a so-called knowledge economy, this is an astonishing self-own.

What we have, then, is a system that:

  • equates institutional affiliation with epistemic competence,
  • penalises open dissemination,
  • and delays circulation until relevance decays.

All in the name of rigour.

I will keep submitting elsewhere. There are other journals. There always are. But let’s stop pretending this is about protecting standards. It is about preserving a hierarchy that mistakes accreditation for insight and treats independent thought as a contamination risk.

Knowledge does not become true by passing through the right doorway. It merely becomes approved. I’ll not witter on about the bollocks of peer review.

That’s Never Happened Before

Famous last words, I know.

I was working with ChatGPT, discussing various concepts. We chatted for several hours over the course of a few days, and we came to, let’s say, a conclusion. What ChatGPT 5 did next was something I had never seen. In fact, I’m sharing the actual response.

I won’t share my thesis here, but this is one of the field verification projects it recommends.

One-page grant prospectus — Provenance Pilot

Project title
Provenance First: Testing provenance packages & slow-mode deliberation to improve municipal decision quality

Lead investigator
Bry Willis — Independent Researcher (contact: [REDACTED])

Summary (1–2 lines)
Test whether mandatory, machine-readable provenance packages plus a 60-day slow-mode and rapid adversarial audits improve decision quality (DQI), reduce reversal rates and increase public legitimacy in municipal planning decisions.

Problem statement (why fund this)
Policy decisions are frequently hijacked by headline pressure and low-provenance inputs; municipalities lack lightweight, testable protocols that make inputs traceable and give deliberation time. This pilot operationalises the “provenance-first + slow-mode” institutional fix and measures whether design moves increase robust decision outcomes rather than performative actions.

Objectives

  1. Test whether provenance packages + 60-day slow-mode increase Decision Quality Index (DQI) vs business as usual.
  2. Measure effects on Provenance Completeness Score (PCS), reversal rates at 12 months, and public legitimacy.
  3. Produce an open replication pack (data, code, provenance schema) and practitioner guidance.

Design & methods (12 months)

  • Setting: medium-sized city planning board; 12 eligible planning items over 12 months.
  • Randomisation: items randomly assigned (stratified by leverage) to Treatment (provenance package + 60-day slow-mode + funded rapid audit) or Control (standard process).
  • Intervention: standard JSON provenance package template (raw data links, code where relevant, funders, changelog), public portal posting, 60-day live comment + one adversarial rapid audit (48–72 hrs) on contested claims.
  • Primary outcome: Decision Quality Index (DQI) — composite of: evidence completeness, process reversibility, and independent expert accuracy score (pre-registered rubric).
  • Secondary outcomes: PCS, reversal at 12 months, public trust (survey), time to decision, stakeholder satisfaction.
  • Analysis: pre-registered mixed effects model (item-level with panel-level random effects), intention-to-treat.

Deliverables

  • Pre-registered protocol on OSF; open data & replication pack (anonymised where required); policy brief and practitioner playbook; 1 peer-reviewed methods paper.

Success criteria (pre-registered)

  • Cohen’s d≥0.4 improvement in DQI for treatment vs control; PCS improvement >25 percentage points; no more than 30% increase in unacceptable delay (pre-agreed threshold).

Risk & mitigation

  • Delay risk: cap slow-mode at 60 days; urgent items can request expedited review (documented override).
  • Capture risk: audit funding ring-fenced and administered by independent trustee panel.

I’m just curious now. Have you ever had a generative AI process end with a cost of goods and services?