20260621
#218
“You can see the computer age everywhere but in the productivity statistics.” — Robert Solow, 1987, quoted in Exponential View
“Systems pushed too far in one direction eventually summon their opposite.” — Why Is This Interesting?
Society: Singapore’s Singlish, creole the state tried to suppress, now a national identity marker. Enantiodromia: internet maximalism about to summon its opposite. Post-reality aesthetics.
Planet: AI for ecological defenders. Mining that looks like farming: the slow-dawning model shift.
AI: Token volumes up 17,000× in four years as inference costs collapsed. Only 15–20% of agentic token consumption is real inference. LLMs as fuzzy compilers. AI prose as meaning-shaped attention vampires.
AI strat: Individual gains don’t compound — “1+1+1+1 = 1.5” at the firm level. The J-curve of general-purpose tech: electrification took 40 years to show up. Agentic coding is a trap. Post-AI employment concentrated in a tiny sliver.
Business: 25 years of startup punditry failure. The AI-era bias toward action — Kahneman’s System 1 risk at organisational scale. Consent beats consensus in fast-moving organisations.
Security: Fuel-tank readers at US gas stations breached. Chatbots know more about you than you realise. The quiet rise of emotional surveillance.
Tech: Uber turning its drivers into a sensor grid for autonomous vehicles. $300,000 robot dogs now patrolling large data centres.
Foresight: Les trois horizons compressés — stabilisation, innovation, rupture, désormais simultanés (FR). What will be scarce in the post-commodity future of work.
Random: Leaf sheep slugs is the animal that photosynthesises on stolen chloroplasts. Iconic restaurants taking reservations for 2046. Keith Haring’s to-do lists. US federal courts and their font question: Times New Roman. Wrigley’s Fascinating Artificial Flavor.
The EU OpenSource Strategy.
,_Anilao,_Filipinas,_2023-08-23,_DD_113.jpg){kind=link}
The Residual
Derek had prepared seventeen slides for the AI Value Summit, which was the kind of name that felt serious until you said it aloud. The slides documented a productivity paradox in careful sequence: his firm’s engineers had submitted 40% more code in Q1. They had closed 60% more tickets. Pull requests had doubled. And yet the organisation had not shipped proportionally more of anything that mattered. He had prepared a graph for this — a slightly baffling bar chart he thought of privately as the 1.5 Slide.
“One plus one plus one plus one equals one and a half,” he said to the room, and paused, as if someone might disagree. The room did not. Avery, the consultant they’d flown in, nodded in the slow way of someone who had seen the same graph in different colours at different companies across an entire industry that had spent three years measuring input and calling it output.
“Lightbulb stage,” she said. “The factory got better lighting. Nobody changed the factory.”
Hank, officially there as head of data science but having spent the morning thinking about sea slugs, looked up. He had read, somewhere in the week’s accumulation of newsletters, about an organism that stole chloroplasts from algae, incorporated them into its own tissues, and ran briefly on solar energy. The slug never actually learned to photosynthesise; it borrowed the machinery and trusted the light to last.^1
“We’re doing kleptoplasty,” he said.
Nobody asked him to continue but he did. “We’re incorporating AI into existing bodies. We haven’t changed the bodies. We’re borrowing the mechanism and hoping the efficiency diffuses.”
Derek considered this. He had forty minutes left in his slot, and an action plan that was basically “do more of the same faster,” which he now suspected was precisely the wrong lesson.^2
Matt, from Infrastructure, interrupted without quite meaning to. He had pulled something up on his phone — a facilities dashboard accessible through the hotel’s sensor API, which had not been configured with authentication. The dashboard tracked something called an Ambient Affect Index for each conference room. This room was reading 62 (Mildly Tense) at the moment of the 1.5 Slide, and had since dropped to 48 (Disengaged).
“The hotel is measuring how we feel,” Matt said.
“Hotels have done comfort profiling for years,” Avery said.
“This is a corporate conference centre. This data goes somewhere.”
Avery looked at the number. 48: Disengaged. She thought of a tipping-point study she half-remembered — 25% of a group, certain enough, could redefine the group’s collective understanding. The question was what understanding she was in 25% of.
“We usually think of these horizons as sequential,” she said, addressing no one in particular. “Stabilise first, then innovate, then transform. But what we’re actually in is all three simultaneously. The infrastructure is being disrupted while it’s being built. The transformation is happening to systems still looking for a stable floor.” She gestured vaguely at the sensor dashboard. “And the building is monitoring our feelings about it.”^3
Derek closed the 1.5 Slide. He had not resolved the graph’s central claim — that individual gains, summed across hundreds of engineers, didn’t compound into organisational results. The explanation wasn’t in the data. It was in the invisible work the data didn’t capture: context re-read forty times; the overhead that looked like inference but wasn’t; the residual that got called productivity because no one had given it another name.
Outside, through the glass, the day’s other events moved in both directions: a wedding reception, a wellness retreat, a seminar on navigating uncertainty. None of them, Hank noticed, were running on their own energy.
^1 The leaf sheep sea slug (Costasiella kuroshimae) can photosynthesise for weeks using stolen algal chloroplasts. Ecologists call this kleptoplasty. Enterprise architects do not use this word, but they might find it useful.
^2 “Do more of the same faster” is Stage 1 of general-purpose technology adoption, corresponding to improved lighting in the old factory. Stage 2 requires restructuring the factory. Stage 3 takes approximately forty years and a generation of managers who don’t remember the previous arrangement.
^3 The three-horizons model — stabilise, innovate, transform — was designed as a planning sequence. Its compression into a single present moment is, depending on who you ask, either a crisis of coordination or an exciting opportunity. The consulting industry charges differently for each interpretation.