LLenny's Podcast
← All episodes
Alex Komoroske (Stripe, Google)03 October 2024

Thinking like a gardener not a builder, organizing teams like slime mold, the adjacent possible, and other unconventional product advice

10Frameworks
16Insights

Frameworks in this episode

Communication5 steps

Cross-Subgraph Resonance Testing (Metaphor Forging)

Test framings live across unconnected audiences; the ones that land in distant subgraphs will travel.

Read →
Strategy5 steps

Gardening Over Building (Farming for Miracles)

Plant many cheap compounding seeds instead of manufacturing outcomes with brute effort.

Read →
Leadership5 steps

Organizational Kayfabe

Diagnose the shared fiction everyone knows is false — then route around it without detonating it.

Read →
Self-Mastery5 steps

Play Yourself Like a Fiddle (Activation Energy + Always-Rules)

Beat activation energy with tight time boxes and an escalating task ladder; hold habits with always-rules.

Read →
Leadership4 steps

Slime Mold Org Design (Sports Car vs Big Rig)

Coordination cost grows with the square of headcount — drive the vehicle you actually have.

Read →
Innovation6 steps

Strategy Salons (Nerd Clubs)

Seed a small, opt-in, yes-and idea group that spins off strategy insights as a side effect.

Read →
Leadership4 steps

Superpower Naming (Be More, Not Different)

Name someone's superpower early, then frame every nudge as 'be more' rather than 'be different'.

Read →
Leadership4 steps

The 70/30 Cover Fire Allocation

Spend 70% earning the right to exist so the other 30% can plant seeds nobody will kill.

Read →
Strategy5 steps

The Adjacent Possible + Low-Resolution North Star

Take only safe next steps — but always the one with the steepest gradient toward a 3-5 year North Star.

Read →
Productivity5 steps

The Bits and Bobs Idea Distillation Loop

Capture everything, filter weekly by resonance, distil, publish — a compounding personal insight engine.

Read →

Insights & moments

The myth-busts, hot takes, explainers, and tools worth keeping.

Hot Take· 4

Hot Take13:30

If Your AI Product Occasionally "Punches the User in the Face," It's Not Viable

Alex critiques teams who treat LLMs as reliable oracles and try to grind down their failure rate. Even at 99% accuracy, an unpredictable 1% that badly hurts the user makes for a non-viable product. Instead of chasing full accuracy, he argues you should design products that assume the tool is 'squishy' and take LLMs for granted as capable-but-imperfect.

  • Reducing failure from 5% to 1% still isn't enough if the failure is severe
  • LLMs are a 'squishy computer' that does roughly what you meant, not exactly
  • Design around unreliability rather than trying to eliminate it
  • Ask what you can build now that you have 'magical duct tape', not how to make it perfect

“even if you get it down to like 99% of the time it's fine if it punches in the face that's not a viable product”

Alex Komoroske · 13:30
#ai#product-design#reliability
Hot Take16:30

In a World of AI Slop, Taste Is the Most Important Skill

As the cost of producing information collapses and most output becomes slop, Alex argues the way to stand out is taste — a perspective distinct from the background average that others find compelling. He defines taste concretely as how different what you say is from what an LLM would have written given the same prompt, and urges people to lean into being the best version of themselves rather than an efficient cog.

  • Falling cost of information production means a cacophony of slop
  • Taste = a differentiated perspective that is both individual and compelling to others
  • Concrete test: how distinct is your output from what an LLM would generate?
  • Good taste must resonate with others, not just be 'out there'

“most of it is slop and so in this cacophony how do you stand out you stand out by having good taste I think taste…”

Alex Komoroske · 16:30

“differentiate from what the llm would have written if given the same prompt how different how distinctive is what you have to say”

Alex Komoroske · 17:00
#ai#taste#differentiation#career
Hot Take31:30

Think Like a Gardener, Not a Builder

Alex contrasts the 'Builder' mindset — make a plan and force reality to match it — with a gardening mindset that looks for things that can grow on their own with a little direction or extra energy. The builder's ceiling is the effort you put in; the gardener plants many cheap seeds and waters the ones that start growing. Done well it looks like magic or luck, because you're 'farming for miracles' rather than picking winners in advance.

  • Builder mindset can't create more value than the effort you put in
  • Gardening finds things with compounding potential that grow on their own
  • Plant many cheap seeds; don't try to predict which becomes the oak tree
  • Water only the ones showing signal; the downside is just the opportunity cost of planting

“what I look for instead are things that can be gardened things that can grow on their own”

Alex Komoroske · 31:30

“what you're doing is you're you're farming for miracles”

Alex Komoroske · 32:30
#strategy#product#emergence#mindset
Hot Take1:20:00

Software Is Alchemy — Don't Let AI Trap Us in a Dozen Boxes

Alex laments that the canonical 'piece of software' has become the monolithic, non-composable app — allowed to exist only if a handful of giant companies permit it. He worries the default AI future is everyone locked in a box with a super-good 'AI clippy', arguing over whose clippy it'll be. He wants the opposite: use LLMs as magical duct tape to escape the box and restore software's combinatorial, agency-extending potential.

  • The default 'canonical software' is now a monolithic, non-composable app
  • Apps exist only if a few of the world's largest companies allow them
  • The feared AI default: everyone locked in a box with a proprietary 'clippy'
  • Software is alchemy — it should combine with others' work in unforeseen ways
  • LLMs as a disruptive force could break the monolithic, aggregator-controlled paradigm

“to me software is alchemy it's the ability to extend human agency Beyond ourselves to create something that can then combine with what others have…”

Alex Komoroske · 1:21:00

“now with the power of AI everyone just is default assuming that what's going to happen is we're all going to be locked inside of…”

Alex Komoroske · 1:21:30
#ai#software#aggregators#web#agency

Explainer· 5

Explainer11:00

LLMs Are "Magical Duct Tape" That Breaks Old Assumptions

Alex describes LLMs as a genuinely disruptive technology that undermines the industry's core assumption that software is expensive to write and cheap to run. LLMs make writing 'good enough' software far cheaper, but their inference cost makes them relatively expensive to run at the margin. He uses the metaphor of a room tilting five degrees: everything looks the same but the direction of gravity has changed, and a lot of your intuition is now wrong.

  • The industry is applying end-of-era playbooks to a technology that doesn't fit them
  • LLMs make writing cheap software cheaper, but inference makes running it costlier
  • Ad-supported consumer startups may not work because ads can't clear inference cost
  • A disruptive tech invalidates assumptions you didn't know you were making

“to me llms are magical duct tape they are formed principally by the distilled intuition of all of society into a thing that operates between…”

Alex Komoroske · 11:00
#ai#llms#disruption#product
Explainer19:00

AI Makes Individuals Better — Which Is Why It's Invisible to Orgs

Unlike most workplace tools that are collaborative, AI makes individuals better in ways they may not want to reveal to their manager. Because it's 'magical duct tape' that's hard to scale into repeatable large systems, it gets used in the small, below the organization's level of awareness. That's why AI can look unused at the industry level while actually being used heavily in the long tail.

  • Most adopted tools are collaborative; AI is the opposite — it boosts the individual
  • People may hide productivity gains for fear of headcount cuts
  • It's easy to help one task, hard to build scaled repeatable systems from it
  • Long-tail usage makes AI's value invisible in aggregate industry data

“this is things that make individuals better and in a way they might not want to tell their manager about”

Alex Komoroske · 19:30
#ai#organizations#productivity
Explainer23:30

Organizational Kayfabe: Why Big Companies Believe Their Own Fiction

Borrowing 'kayfabe' from pro wrestling — something everyone knows is fake but acts as if real — Alex explains how small optimistic white-lies in status updates compound up the org chart until leadership is orders of magnitude off ground truth. Anyone who tries to 'hit the ground truth button' risks being knocked out of the game, so people eventually just believe the fiction, and the organization becomes a zombie that lumbers on despite everyone privately agreeing it can't work.

  • Kayfabe = something everyone knows is fake but treats as real
  • A 'yellow' status quietly reported as 'green' is locally rational but compounds upward
  • The core asymmetry: you can't make your boss look dumb, so bad news gets buffered
  • Eventually people just believe the fiction, turning the org into a zombie
  • Alex insists this is descriptive, not cynical — naming it reveals cheaper options

“kab is a word that comes I believe is a Carney word that is used and applied to professional wrestling and it means a thing…”

Alex Komoroske · 24:00

“you can't make your boss look dumb because if you do they're the person who who decides oh this person's not performing”

Alex Komoroske · 29:30
#organizations#leadership#systems#culture
Explainer42:30

Slime Mold Orgs: Don't Drive a Big Rig Like a Sports Car

Alex's slime-mold thesis is that coordination cost grows roughly with the square of the number of people, making large orgs hard to steer even when everyone is good and collaborative. A founder can steer a small company like a sports car; at scale it becomes a big rig, and driving it like a sports car wrecks it. The alternatives are to drive the big rig deliberately (invest in program management, pivot less) or split into a swarm of autonomous sports cars — Apple chose coherence, AWS chose the swarm.

  • Coordination cost scales with roughly the square of the number of people
  • Small company = steerable sports car; large company = a big rig
  • Big rigs mean 'pivot less' and invest in process, not just 'go slow'
  • Apple optimizes for coherence; AWS embraces an antifragile swarm
  • Slimes can find solutions to problems you didn't know you were searching for

“if you drive a big rig like a sports car you're going to be a danger to yourself and others on the road and you're…”

Alex Komoroske · 43:30

“slimes are actually kind of amazing they can find pro solutions to problems you didn't even know you were searching for”

Alex Komoroske · 45:00
#organizations#scaling#systems#strategy
Explainer1:08:00

The Adjacent Possible: Small Steps Toward a Low-Res North Star

Alex explains 'the adjacent possible' — the small set of near-certain actions right in front of you. Rather than making a flying leap to an end state, you take a safe next step, the world reconfigures, and new options open up. You pair this with a deliberately low-resolution North Star three-to-five years out that everyone agrees is plausible, then repeatedly pick the adjacent step with the steepest gradient toward it — needing both incremental steps and long-term coherence to avoid random-walking or building castles in the sky.

  • The adjacent possible is smaller than we assume — within arm's reach
  • Each safe action reconfigures the world and unlocks the next set of options
  • Set a low-resolution North Star 3-5 years out that everyone finds plausible
  • Pick the adjacent step whose gradient pulls hardest toward the North Star
  • Incremental-only random-walks; long-term-only builds impossible castles

“when you recognize that your adjacent possible is relatively small you realize that you actually have full agency to pick within the subset that is…”

Alex Komoroske · 1:08:30

“the way you get that is by creating a kind of North Star for yourself it should be in three to five years in the…”

Alex Komoroske · 1:10:00
#strategy#product#planning#decision-making

Story· 1

Story02:00

The "Bits and Bobs" Doc: Alex's Idea-Collecting Practice

Alex keeps a ~600-page public Google Doc plus 17,000+ private working notes where he captures interesting ideas from conversations every day. During the week he processes and tags notes (using embeddings to surface similar past ideas), then on Friday afternoons selects the ones that still resonate, distills them on weekends, and publishes on Mondays. He frames this reflective writing habit as the single biggest source of his most interesting insights and a driver of his productivity.

  • The public 'bits and Bobs' doc is roughly 600 pages of unspooled insights
  • He has 17,248 unpublished working notes in an open-source tool built ~5 years ago
  • Uses embeddings to find similar past ideas when processing new notes
  • Weekly cadence: capture daily, select on Friday, distill on weekends, publish Monday
  • Deliberately keeps the doc a little 'illegible' so readers have to work with it

“it's like I can't imagine not doing this it is the the place where I find most of my most interesting insights is by reflecting…”

Alex Komoroske · 07:00
#note-taking#productivity#reflection#writing

Tool· 1

Tool20:30

Using Claude as an "Electric Bike for Idea Spaces"

Alex says he talks to Claude ~20 times a day, loading projects with context from his bits-and-bobs notes and using it to think through problems — generating examples, critiquing framings, and layering in concepts. He frames it as a well-read but naive friend who never makes you feel dumb, and updates Steve Jobs's 'bicycle for the mind' metaphor: LLMs are an electric bicycle for the mind, letting you cover far more idea-ground quickly.

  • He talks to Claude ~20 times a day with heavily context-loaded projects
  • Uses it to name concepts, generate 10 examples, and critique his own framings
  • Think of it as an extremely well-read but naive friend who won't judge you
  • Explore a problem space cheaply, then check conclusions with real experts

“I I talk to Claude 20 times a day”

Alex Komoroske · 20:30

“it's like an electric bike for idea spaces kind of you can just cover so much more ground so much more quickly in them”

Alex Komoroske · 22:00
#ai#claude#tools#thinking

Takeaway· 5

Takeaway08:00

Take a Reflection Day: Deep Thinking Needs Space

Alex argues that being constantly busy leaves no room for the deep thinking that actually saves time. He keeps one meeting-free day (originally Fridays working from home, despite colleagues mocking him) to read carefully and reflect. His test: if he had to explain the same strategic idea in ten one-on-ones, that should have been a 30-minute document instead.

  • "The Mundane Pointless will cut take every square inch you give it" — so you must protect space
  • Monday-Thursday is meetings 8-to-6; one day reserved for reading and reflection
  • If the same framing worked for 10 people, write it as a reusable document
  • He does this for effectiveness, not just enjoyment

“deep thinking takes time and space and you got to create that space”

Alex Komoroske · 08:00
#productivity#deep-work#time-management
Takeaway35:30

Create "Cover Fire" So Your Team Can Do Great Work

Alex's approach at Google was to have 70% of his team's effort go toward clearly valuable, even boring, linear work — so no one could ever ask 'what does that team even do?' That undeniable value buys credibility and protection, freeing the other 30% to plant risky seeds. He pictures it as laying down cover fire while quietly building the 'doomsday bomb' where no squirrel can dig up the acorn.

  • 70% of effort on unambiguously valuable work to earn organizational trust
  • The remaining ~30% funds experimental, hard-to-measure seeds
  • Credibility is what protects a team's space to do great (not just good) work
  • Great work requires a high-trust environment where people lean into superpowers

“you want to create cover fire for your team where your team's just hitting goals moving metrics and then while with that cover fire you're…”

Lenny Rachitsky · 37:30

“70% of my effort and my team's effort should go on things that everybody acknowledges are important and useful and create value”

Alex Komoroske · 36:00
#leadership#teams#strategy#management
Takeaway57:00

"Always Rules" Beat "Sometimes Rules" for Self-Control

Alex's productivity insight is that clear black-and-white rules hold up better than negotiable ones. A 'sometimes' diet rule collapses the first time you rationalize an exception; an 'always' rule (with narrowly-defined exceptions) is easy to hold forever. He's done a Peloton workout every single day since the pandemic began by asking, of any given day, whether it's really the single hardest of hundreds — it never is.

  • Sometimes-rules break the first time you rationalize an exception
  • Always-rules are clear enough to hold indefinitely
  • Define narrow, unambiguous exceptions (e.g., a socially-awkward homemade dessert)
  • The daily-streak framing ('is today really the hardest day?') sustains the habit

“always rules are better than sometimes rules for self-control”

Alex Komoroske · 57:00

“is today the day of all the hundreds of days in this streak that is the worst or the hardest for me to do this…”

Alex Komoroske · 58:00
#productivity#habits#self-control
Takeaway1:03:30

The Hallmark Card Fallacy: Why Clichés Are Actually True

Alex describes the 'Hallmark card fallacy': you earn a deep insight at great effort, try to share it, and people dismiss it as a greeting-card cliché. The reason those phrases get repeated endlessly is precisely that they're meaningful — you just heard them before you were ready. He tries to plant ideas like little seeds that may only crack open years later when someone's finally ready.

  • Hard-won insights often sound like trite clichés to others
  • Clichés are repeated so often because they're genuinely meaningful
  • You can hear the words a million times but only 'get' them when ready
  • Plant ideas as seeds that may germinate in someone years later

“the reason people keep on saying it is because it's meaningful”

Alex Komoroske · 1:04:00
#wisdom#communication#mentorship
Takeaway1:13:30

Happiness = Reality Minus Expectations

Riffing on Tim Urban's formula, Alex notes that reality is hard to change but expectations are easy, so holding expectations lightly is the cheapest lever on happiness. He's careful to say this isn't an argument for mediocrity: still seek something great and be proud of it, but don't rigidly attach to a specific outcome. A companion tip: feel now the emotion the story will give you in ten years (if it'll be funny later, find the humor now).

  • Reality is hard to change; expectations are easy, so adjust expectations
  • Hold expectations lightly rather than lowering ambition
  • Not an endorsement of mediocrity — still seek something you're proud of
  • Feel the emotion your future self will have about the story (funny-in-10-years)

“happiness is um reality minus expectations”

Alex Komoroske · 1:13:30

“reality is hard to change it's not impossible it's hard to change your expectations are super easy so just change your expectations”

Alex Komoroske · 1:14:00
#happiness#mindset#life-advice