Gardener thinking, slime mold teams, and navigating AI disruption with Alex Komoroske

Executive overview

Most product thinking is builder-brained: make a plan, execute it, cap your upside at the effort you put in. A gardener mindset does the opposite — plant cheap seeds, respond to what grows, let compounding do the work.

LLMs are a genuinely disruptive technology that tilts the room five degrees. Intuitions built in the old era are now subtly wrong, and the teams that win will be the ones who play rather than pattern-match.

The highest-leverage skill in an AI-saturated world is taste — a perspective distinctive enough to stand out from the background noise LLMs generate by default.

Gardening versus building

• Builder mindset caps value creation at effort invested; gardener mindset seeks compounding, self-accelerating dynamics.
• Look for ecosystem shapes: network effects, compounding loops — anything that grows faster once it starts.
• Plant cheap seeds; invest incrementally only when you see signal someone finds it useful.
• Stop investing if signal disappears. Never commit to a seed before it sprouts.
• Protect the acorn: a leader's credibility buys the space for the team's best ideas to survive long enough to grow.
• 70% of effort on clearly legible value creation; 30% freed for seeds the org can't yet see.

LLMs as disruptive technology

• LLMs are "magical duct tape" — distilled human intuition at a cost between a human and plain computing.
• They undermine two core software-era assumptions: expensive to write, cheap to run.
• The room has tilted five degrees — gravity now pulls differently, and old intuitions produce wrong outcomes.
• Consumer AI products can't be ad-funded; inference cost doesn't clear advertising revenue.
• Design for squishiness: a product that punches users in the face 1% of the time is not viable, regardless of average performance.
• Most AI adoption is invisible to organisations — individuals gain 2x leverage and don't announce it.

Skills that matter most now

• Curiosity and play over execution playbooks — we're in the community gardening phase, not factory farming.
• Taste: what you produce must be distinguishably different from what the LLM would write given the same prompt.
• Lean into your genuine perspective rather than optimising to be a better cog.
• Seek the weird, alien, and compelling (e.g. WebSim) — these reveal what's only possible in an LLM world.
• Vertical SaaS is not the right frame for AI-native problems.

Organisational kayfabe

• Kayfabe: something everyone knows is fake but acts as if it's real — the lubricant that keeps organisations moving.
• Optimism compounds through reporting layers; ground truth degrades many orders of magnitude by the time it reaches leadership.
• The underlying asymmetry driving it all: you can never make your boss look dumb.
• This is emergent — it shows up even when everyone hates it and no one chose it.
• Fighting it without acknowledging it produces zombie organisations where individuals privately agree nothing will work.
• Solution: create low-stakes channels for disconfirming evidence to arrive continuously rather than in one shattering moment.

Slime mold organisations

• Coordination cost grows roughly with the square of team size — this is inescapable.
• Small company = sports car; large company = big rig. Drive the vehicle you actually have.
• Two valid architectures: tight coherence (Apple) or autonomous swarm (AWS). Both work; pick deliberately.
• Autonomy-first tech companies are already more like slime molds than they realise — embrace it rather than fight it.
• Slime molds find solutions to problems they weren't explicitly searching for.

Strategy salons and nerd clubs

• Create optional, secret, low-stakes groups where people participate purely because they want to.
• Set norms explicitly: "yes, and" only — disagreement is expressed as "I wonder if…" not "that's wrong."
• Dribble in new perspectives one to three at a time; bulk additions scramble norms.
• Diversity of background in who resonates with an idea predicts ceiling of broader reach.
• Feed momentum: ensure there's always quorum; send FOMO notes after live sessions.
• The group surfaces ideas organically; never steer it toward a predetermined outcome.

The adjacent possible and North Stars

• Adjacent possible: the set of actions actually within reach right now — smaller than people assume.
• Each action reconfigures the world and opens a new set of adjacent actions.
• Incremental-only thinking random-walks you into a corner; you need a North Star to arc toward.
• North Star: 3–5 years out, low resolution, plausible to every stakeholder, worth high-fiving if reached.
• North Star should update slowly (sliding, not jerking) as new information arrives.
• At each step, pick the option with the steepest gradient toward the North Star.
• False precision in forecasts is a comfort blanket, not analysis; compounding ideas produce order-of-magnitude variance, not 93 vs 95.

Productivity and self-knowledge

• Deep thinking requires manufactured space — mundane work fills every inch you give it.
• Always rules beat sometimes rules for self-control: a clear bright line holds indefinitely; an exception-riddled rule breaks the moment pressure arrives.
• Activation energy is real: structure work so small completions build momentum into harder tasks.
• Capture ideas immediately — a butterfly collector mindset; delayed capture loses them.
• The compendium habit: fast notes during the week, weekly reflection pass, publish on a rhythm.
• Core philosophy: do things that give you energy and that you are proud of — this makes effort intrinsically rewarding and eliminates the hidden cost of shortcuts.