Principle · MindFinnish soldiers taught Diamond vocabulary by picking up a cup and naming it (kupi), and demonstrated postpositional case endings by moving cups onto and off of tables; Diamond recommends learning from native speakers directly rather than only from textbooks.
anecdotal
Andrew Huberman · What Makes Great Leaders & How to Live an Extraordinary Life | Dr. Jared Diamond
Principle · MindDiamond maintains his Italian by doing all leisure reading in Italian and having a two-hour weekly conversation with a native-speaker tutor, a pattern the host also notes as a couple of hours per week.
anecdotal
Andrew Huberman · What Makes Great Leaders & How to Live an Extraordinary Life | Dr. Jared Diamond
Principle · MindWhen Diamond considered dropping out of science, his father gently advised that it might be a bit early since he had been doing it for only a year, and suggested giving himself another six months before making the final call; Diamond endorses waiting at least a year before quitting.
endorsed
Andrew Huberman · What Makes Great Leaders & How to Live an Extraordinary Life | Dr. Jared Diamond
PrinciplePamela Fox recommends starting with the most low-level tool calling you can do to understand the fundamentals, before abstracting to agents. For building a proper agent, use an agent framework — picking from options like LangChain, Pydantic AI, and Microsoft Agent Framework.
AI Engineer · The Model Swap Workshop: GPT-5.5 vs Kimi vs DeepSeek vs Mistral — Pamela Fox, Microsoft
Creator perspective · MoneyNaidu describes the typical path for an AI vendor selling into the enterprise: find a pain point, build an agent for a narrow use case, then face security review, internal coordination, and integrations with 20-year-old systems — the 'first wall' of deployment.
AI Engineer · Why Buying Companies Beats Selling Them AI — Prathik Naidu, Long Lake
Principle · MoneyNaidu argues that even if a vendor delivers end-to-end value, capturing that value is hard: landing a seven-figure contract raises questions about whether customers will keep renewing and whether pricing can be maintained as models improve and alternatives get cheaper. As a vendor, you do all the deployment and integration work, but the largest impact shows up in the client's P&L — a $5M/year EBITDA impact at a conservative 10x multiple generates $50M in enterprise value for the client, not the vendor.
AI Engineer · Why Buying Companies Beats Selling Them AI — Prathik Naidu, Long Lake
Creator perspective · MoneyServices businesses face the challenge that labor costs grow with revenue: a community manager can only run so many associations, an architect can only design so many hospitals or schools. In software, $100 of revenue keeps about $80 due to low marginal cost; in a service business, $100 of incremental revenue typically carries around $80 of delivery and overhead because you have to hire, train, and manage your team. Many founders have stopped chasing growth — not from lack of ambition or market opportunity, but because of these scaling challenges. Naidu frames this as an 80% marginal tax on growth that makes these businesses hard to start and scale.
AI Engineer · Why Buying Companies Beats Selling Them AI — Prathik Naidu, Long Lake
Creator perspectiveFin-Syn being repealed is one of the worst things to happen to the TV industry in the last several decades, because networks started picking only from in-house content, which means bad stuff gets out there.
Acquired · Seinfeld
Creator perspectiveJerry and Larry's deal gave them huge back-end from the production company, letting them make the show artistically what they wanted while being incentivized for the long run.
Acquired · Seinfeld
PrincipleWhen a partner hesitates on a big commitment, offer a small trial order instead of a full one — NBC ordered just four episodes, the smallest first season order in television history.
Acquired · Seinfeld
PrincipleJeff Schaefer's test for Seinfeld material: the plot itself has to be funny when you retell it — if literally describing the situation is the humorous part, you're onto something. The narrators generalize this: test creative work by retelling it aloud; if the summary itself isn't compelling, the underlying idea needs work.
Acquired · Seinfeld
Creator perspectiveWhen the network pushed back, Larry David blew up at Rick, threatening to quit if they didn't make the show his way. The hosts note that no creator before Larry David would threaten to quit when the network disagreed with their vision.
Acquired · Seinfeld
Creator perspectiveThe DNA of Seinfeld is the stand-up callback structure — the show is built like a great stand-up routine, where at the end you realize everything was interwoven and building to that moment. Seinfeld also threw out the act structure — plots often don't resolve. Structuring interwoven story is hard: you must introduce elements without spoiling the place where storylines converge, then pay off at the right time.
described
Acquired · Seinfeld
ReportSeinfeld was born when Jerry and Larry joked about products on the shelves at Lee's Market, a Korean deli near Catch a Rising Star.
Acquired · Seinfeld
ReportIn season four Larry and Jerry went on an all-timer of a season, starting the first season-long meta arc that became a show-long meta arc. In it, a fictional NBC asks Jerry to pitch a TV show called Jerry, and he and George come up with the idea of a show about nothing to pitch to NBC. The meta arc runs off and on throughout the whole season, and the off-and-on is important because each episode still stands on its own — you don't have to be invested in the multi-episode arc to get it.
Acquired · Seinfeld
ReportTypical sitcoms use an A story taking 60-70% of the runtime plus a separate B story they flip between, and a typical two-act sitcom has seven scenes in 22 minutes. Seinfeld instead gave each of the four characters their own concurrent storyline, requiring extremely efficient storytelling with no wasted words or minutes. Storylines start interwoven in unexpected and surprising ways as episodes progress, then converge at the end in a violent, unexpected blow-up that's unfortunate for everyone, then cut to credits. In Seinfeld, no situation is there for filler — nothing is wasted.
Acquired · Seinfeld
PrincipleStructure your work like a great stand-up routine: plant elements throughout that all interweave and build to a single climactic payoff, with nothing there for filler. When running multiple concurrent storylines, be ruthlessly efficient — no wasted words, no wasted minutes. When structuring an interwoven story, introduce elements early, pay them off later, and avoid spoiling the convergence point in the middle — a specialized, difficult skill. Seinfeld's season-long arcs were designed so each episode still stands alone.
Acquired · Seinfeld
Creator perspectiveCastle Rock's business model was to own the forever rights to all downstream revenue after first-run television, in the Silicon Valley style of independent production companies betting on long-term equity value — like venture capital.
Acquired · Seinfeld
ReportEmmanuel Ameisen explains the scaling law observation — pouring more compute and data into next-word prediction predictably improves the model's skill in a continuing trend. The scaling loss paper came from the early Anthropic team building on an OpenAI observation, and the law remains mostly empirical, with people speculating about why.
Also discussed by Joseph Kim on AI Engineer (unclassified source — not counted until classified). Not counted as independent sources.
Lenny's Podcast · Inside the black box of AI: How AI models actually think | Emmanuel Ameisen (Anthropic)
PrincipleEmmanuel Ameisen argues usefulness depends on accuracy: at roughly 90% of medium words right, a model is useful as an email writer; at 50% accuracy it wastes your time because you must fix it half the time. Different tasks have different reliability thresholds at which the model becomes useful.
hypothetical framing by speaker
Lenny's Podcast · Inside the black box of AI: How AI models actually think | Emmanuel Ameisen (Anthropic)