Startup Anti-Patterns

Itamar Novick (Recursive Ventures) and Simeon Simeonov's series intro on startup anti-patterns — named, repeatable patterns of failure that look like sensible moves but stack into poor outcomes. The thesis: failure is more legible than success, so study what kills startups rather than what makes them win. The post is mostly a vocabulary dump — ~70 named anti-patterns — waiting to be filled in by later essays.

What It Is

An anti-pattern is a commonly used process, structure, or pattern of action that, "despite initially appearing to be an appropriate and effective response to a problem, has more bad consequences than good ones." The term is borrowed from software (where an anti-pattern is a known-bad design), then widened to cover every decision a startup makes — hiring, fundraising, strategy, product, psychology.

Simeonov first made the case in 2013; the series relaunches it with Novick as co-author, promising "tangible examples from our experiences as founders and investors in 100+ startups, and the experiences of guest founders from our portfolio."

Key Quotes

"Studying repeatable patterns of startup failure (startup anti-patterns) is more useful than studying non-repeatable strategies for startup success."

The thesis in one sentence, and the sharpest claim in the piece. Success is over-determined and idiosyncratic — the exact recipe that worked for one company won't transfer. Failure is legible: the same handful of mistakes recur across companies, so a taxonomy of them is a reusable negative checklist. It's the same move A Pattern Language (Christopher Alexander) makes for buildings, inverted — name the patterns that wreck an environment rather than the ones that make it flourish.

"Pursuing an anti-pattern doesn't mean that your company will die tomorrow or in the next year, but each anti-pattern adds-up and could lead to clouding your focus and hampering your ability to execute."

The authors pre-empt the obvious objection. Anti-patterns aren't fatal on contact — they're slow leaks. The danger model is accumulation: one anti-pattern you can absorb, five you can't, and the damage shows up as diffuse fog ("clouding your focus") rather than a single legible cause of death. This is why the list form matters — the point is to notice how many you're running at once, not to avoid any single one.

"it's hard to pinpoint the exact set of reasons startups succeed, but experienced entrepreneurs and investors have a good sense of what drives startups' failures."

The epistemic asymmetry underneath everything: failure is observable and common, success is rare and noisy. A founder can't reverse-engineer Stripe, but they can avoid the mistakes the authors have watched kill a hundred other companies.

Key Themes

#concept Anti-pattern as the mirror of a pattern language — Where Alexander's patterns prescribe conditions for human flourishing, anti-patterns catalogue conditions for failure. Same grammar (named, recurring, composable), opposite sign.

#concept Failure is repeatable; success is not — The load-bearing claim. It justifies the entire enterprise and quietly argues against the survivorship-bias literature that treats founder stories as transferable playbooks.

#concept Risk accumulation, not single-cause doom — Anti-patterns compound. The failure mode is fog, not a cliff: focus clouds, execution slows, and the company dies of a thousand micro-decisions rather than one big one.

#pattern The taxonomy as negative checklist — The ~70-item list is the real artifact. Elephant hunting, Platform risk, If you build it they will come, Bad revenue, Analysis paralysis, Founder arrogance, Boiling the ocean, Design by committee, Confirmation bias, Premature optimization, Featuritis, Founderitis, Groupthink, Death by pivot, Overengineering, Unicorn hunting, Zombie. Read as an audit tool, not a theory.

#person Simeon Simeonov — the originator (2013) of the startup-anti-pattern framing. #person Itamar Novick — Recursive Ventures; co-author bringing the series back with portfolio-founder case studies.

Critical Analysis

The strongest idea is the inversion of the success-porn genre. Most startup content — including The Founder's Playbook — tells you how to win. This tells you how to not lose, which is a categorically different and more tractable ask. You can't copy another company's win; you can avoid the failure mode that already has a name.

But it's an introduction, not an argument. The post is a thesis statement plus a list. Every anti-pattern is a bare noun phrase with no definition, example, or severity weighting — the authors even flag that the list is "not drawn to scale," with some patterns more frequent and more fatal than others, but they don't say which. Until the follow-up essays land, the value is the vocabulary itself: giving founders words for the bad thing they're doing is often the first step to stopping it.

The list is a fascinating Rorschach test. It mixes strategy failure (Chasing Blue Oceans, Bad revenue, Next round only), process failure (Death by pivot, Top-down planning, Design by committee), and plain psychology (Confirmation bias, Groupthink, Arrogance, Escapism, Oooh shiny!). That breadth is the quiet argument — startups die for reasons that are mostly human, not technical. The pattern catalogues map cleanly onto Software Engineering Craft's war on complexity: Overengineering, Premature optimization, and Featuritis are the same named demons whether the thing being built is a startup or a codebase.

The mirror relationship to Moats — 80 Strategies for Competitive Advantage is the most useful cross-read. Moats is a taxonomy of how things win — eighty named, amoral strategies, flat and unweighted. This is the negative space: how things lose. Read together they're a two-sided audit — are you accumulating an edge or accumulating a leak?

What it doesn't touch: the series predates (or ignores) the AI-native failure modes that The Founder's Playbook adds — "confirmation bias with a research engine," agentic technical debt, zero-friction scope creep. "Confirmation bias" is on the list, but the AI twist isn't: when building is nearly free, the old anti-patterns get faster, and the list may need a second edition for the age of agents.

Sources: raw/intro-to-startup-anti-pattern-series, Intro to Startup Anti-Pattern Series Last updated: 2026-09-04