Libido Sciendi Digest #05 - The Frame Decides the Count

Hacking’s looping effects, Subran’s 84% gap, mitochondria as engulfed bacteria, Linnaeus’s 1735 branches: where the frame stops is what counts.

This is your Libido Sciendi Digest. Weekly notes from a book-obsessed seed investor exploring AI, living systems, ethics of technology, emergence, and the diffusion of innovation. Essays, field reports, book notes and curated readings from libido-sciendi.com.

If you enjoy it, I would genuinely appreciate it if you shared it with someone who might too. And if you have feedback on the format, just reply to this email.

#05

Two pieces shipped on the same day, both performing the same operation: take a continuous gradient, choose where to draw the line, and the line becomes the answer.

The frame disappears and decides almost everything that follows. Every classification system tempts us to forget that, and the bill is paid in law, in ethics, or in clinical practice. Psychometrics is my own running discomfort, with DSM thresholds drawn on smooth gradients and then producing the realities of who gets a diagnosis, a service, or a label they will carry for life.

Ian Hacking (1936-2023, Canadian philosopher of science) talked about the looping effects of human kinds: classify people (anorexic, autistic, multiple personalities, “the kind of person who…”) and the classification feeds back onto the classified, who become more like the category, which in turn changes the category, in a loop the original namers do not control. The boundary does not just describe reality, it produces it: the diagnostic category changes the people, the legal category changes the rights, the conservation category changes the species. Hacking’s Rewriting the Soul (1995) and his short essay “Making Up People” (2006) are canonical entry points, and the question runs deep enough into epistemology that I may write a proper journal entry on it before long.

Blaise Agüera y Arcas pulls the same rug from under biology: what counts as alive dissolves into where the frame is held, with mitochondria as two-billion-year-old engulfed bacteria, Daisyworld as a regulating planet, Linnaeus’s 1735 branches as the unstated assumption beneath every welfare law and every conservation priority.

Ludovic Subran and Markus Zimmer point the same lens at capital allocation: Europe's climate transition needs EUR 480 billion more per year, an 84% gap against current flows. But AI capex now competes for the same constrained pool of capital: both pull from the same European savings, the same fiscal space, the same balance sheets. Move the line so AI capex sits inside the same budget constraint and the 84% becomes an allocation question across two priorities; climate transition stops being just a climate financing question.

Reading them in the same week, I had to admit that for decades I have thought I was attentive to these things, suspicious of categories, alert to construction. And yet, like anyone, I used Linnaeus’s branches as if they were the joints of nature, and treated discount rate as if it were arithmetic rather than a politically charged claim about priorities, and whose welfare counts and when. The frames that do the most work are precisely the ones that have stopped feeling like frames.

So an exercise, for today and on some quiet regular cadence after. Pick a category you have been working with for years without question (could be a number, a subset, any boundary). Ask where the line was drawn, by whom, against which alternative, and what would happen if it sat one notch higher or lower. The point is rarely that the line is wrong.

The point is almost always that it was a choice, and you, we, had stopped noticing.

The 84% Gap: What Climate Investment Actually Requires · 24 min read

Part 3 of the AI × Energy series. Subran and Zimmer’s Investing in a Changing Climate gives the cleanest accounting yet of the European climate ledger. The numbers, at a glance:

• The 84% gap. Actual climate investment in Europe runs at roughly 16% of what the transition requires, leaving an 84% gap that compounds every year it is not closed.

• EUR 480 billion a year, the additional capital needed, drawing on the same constrained pool as AI capex.

• Net zero is not the same as 1.5°C compatible. Net zero by 2050 (the EU target) caps the endpoint; warming is determined by the area under the emissions curve. The 1.5°C carbon budget (~400-500 GtCO2 remaining as of 2020 per IPCC AR6) is on track to be exhausted by the early 2030s even if every net-zero pledge holds, because emissions in the 2020s stay too high. The endpoint metric and the budget metric give different answers about whether the path is on track.

• DAC as the closing trade. Direct air capture is the only technology that lets the math close after the budget is overspent, by pulling CO2 directly from ambient air rather than capturing at the smokestack. Current costs run $400-1,000 per tonne; if DAC settles toward Bilal and Känzig’s $1,200-per-tonne damage estimate, large-scale deployment becomes unilaterally rational, but the early CAPEX has no obvious underwriter, and “AI capex versus DAC capex” is the same pool-competition argument at a different scale.

The piece's contribution is also a framing move: putting AI capex in the same frame as climate capex makes the trade-off visible and forces an answer to "who decides which capex line gets cut first when the pool stops growing."

Both have a public-good case no private rate-of-return calculation captures, which is exactly why they compete politically, not just economically. What gets funded is also a story of what gets defunded silently, each quarter as the frame stays the way it is.

Reading What Is Intelligence? (2): Where Does A Life Begin? · 10 min read

Notes on Chapter 1 of Agüera y Arcas’s What Is Intelligence?, where the question of what counts as alive dissolves into a question about where the frame is held.

Functionalism treats substrate as interchangeable, arguing that function alone determines identity, that what a heart does is what makes it a heart. Applied to biology, this idea extends in a somewhat unsettling way across all scales, from entire planets down to individual ribosomes.

Three measurable criteria replace the threshold question of is this alive:

• how much computation,

• how much free energy,

• how much robustness to perturbation.

Linnaeus’s 1735 branches are still the unstated assumption beneath every welfare law and every conservation priority, and the chapter quietly suggests the ethical frameworks we use are choosing what to protect by choosing where to hold the zoom.

If every classification protects something at one scale by implicitly refusing to protect something at another, on what grounds do we still hold the zoom where we hold it?

Ten concepts, five from each entry. Full toolboxes at the end of each article.

• From The 84% Gap:

  • the 84% gap: the climate-investment deficit, with actual investment running at roughly 16% of what the European transition requires;

  • discount rate as politics: the rate used in climate accounting is normative;

  • capital pool competition: AI capex and climate capex now drawing from the same constrained European pool, with no mechanism in place to arbitrate between them;

  • domestic cost of carbon: Bilal and Känzig, the unilateral $216 US / €226 EU per-tonne pricing that decouples decarbonisation from international cooperation;

  • public-good externality beyond IRR: the case both climate and AI capex have that no private rate-of-return calculation captures, which is exactly why they compete politically, not just economically.

• From Where Does A Life Begin?:

  • functionalism: Turing and von Neumann, substrate is interchangeable; identity sits in what the system does, not what it is made of;

  • symbiogenesis: Margulis, evolution’s big jumps come from mergers; Agüera y Arcas extends the frame to human-machine cognition;

  • Daisyworld: Lovelock and Watson 1983, local selection on albedo producing global temperature regulation, Gaia at the planetary scale;

  • the three living-system criteria: Agüera y Arcas, computation + free energy throughput + robustness to perturbation, replacing the threshold question of “is this alive”

  • the Linnaean default: 1735 species hierarchy still the unstated assumption beneath every welfare law, conservation priority, and intuitive moral category about harm.

Seven readings, split across the week’s two scales: four on cognition (where the line gets drawn between thought and pattern-completion, and what happens when that line moves), three on capital (where the line gets drawn between AI capex and everything else competing for the same pool).

Cognition

• The illusion of thinking, The Signal (Alex Banks), 8 min
  The companion question to the journal entry, asked from the cognition side: at what point does pattern-completion stop counting as thought, and on whose criteria.

• The “Cognitive Offloading” Paradox, Dr Philippa Hardman, 12 min
  Hardman extends the cognitive surrender argument: the more you offload, the more confidence you carry about the thinking you did not actually do.

• Nature Medicine Study Shows AI Outperforms Therapists on Cognitive Behavioral Therapy, Business Wire, 5 min
  The clinical case study for the boundary problem. Nature Medicine reports a randomized trial in which AI-delivered CBT beats trained therapists on standardised outcomes. Read after Hardman and you have the full pincer: AI scores higher on the metric, the metric was designed for a category that did not include AI, and the diagnostic frame is what gets revised first.

• The Bitter Lesson, Rich Sutton, 5 min
  The classic counter-claim to the symbiogenesis frame. Sutton argues general methods that scale beat human-engineered structure every time, which is the position Agüera y Arcas’s chapter quietly contests. Worth re-reading before believing either of them.

Capital

• Forging a New AGI Social Contract, Windfall Trust, long read
  The normative-frame layer the deep dive only gestures at. The piece lays out what an AGI social contract would actually contain (ownership of compute, ownership of weights, ownership of the rents), and the answer is the same Subran one in different vocabulary: the rules get drawn before the surplus does.

• Interview with Jensen Huang on Accelerated Computing, Stratechery, long read
  The capital-pool view from the supply side. Huang on why every euro of climate capex now passes through the same datacentre power-and-silicon corridor as every euro of AI capex, told from the company that prices the corridor.

• Dylan Patel: Bottlenecks to Scaling AI Compute, Dwarkesh, long read
  The hard-numbers version. TSMC sold out through 2027, the EUV monopoly setting the cumulative ceiling, HBM demand cannibalising consumer DRAM at 90 to 95% price moves: the supply side of the same pool Subran is trying to redirect.

  

  

  

• Development as Freedom, Amartya Sen, 1999 (★★★★★)
  Sen reframes well-being from GDP to capabilities, the same operation Agüera y Arcas runs on biology and Subran runs on the climate ledger. Recently Read landed it at the top of the list in the slipstream of #04’s IHEMI evening, but it never made the Library section there. The book is a reminder that the frame question is not an exotic philosophical move; it is the discipline-by-discipline replacement of one default with another.

  Behind The 84% Gap
  

• Investing in a Changing Climate, Ludovic Subran and Markus Zimmer, 2023 (★★★★★)

  The book the deep dive is built on. Subran (Allianz’s chief economist) and Zimmer apply the actuarial machinery of insurance to the European climate ledger and arrive at the EUR 480 billion-a-year shortfall by reading the discount rate as a political choice rather than a technical parameter.

The journal entry’s source book, Agüera y Arcas’s What Is Intelligence?, was extensively covered through the symbiogenesis essay in #03 and the Currently Reading note in #02; Chapter 1 is read in full in this week’s entry, so it is not re-featured here.

From a 2-billion-year-old engulfed bacterium to a EUR 480 billion-a-year climate gap, the operation is the same: where the frame stops decides what gets counted, and the day the frame moves the math changes too.

libido-sciendi.com · Subscribe to the Digest