Martin Källström
knowledge / philosophy

Failure & resilience

Martin's clearest thinking on failure comes from a single, specific wound: Narrative, the lifelogging camera company he built for five years and lost to bankruptcy in September 2016. Everything else in this file — the discipline of shipping "ready enough," the instinct to own a problem instead of assigning blame, the conviction that value matters more than success — reads as commentary on that one event. His central claim is almost mechanical: companies don't fail because of bad ideas, bad products, or bad luck; they fail because they run out of cash, and the decision to stop fighting is far more subjective than it looks from the outside.

It always comes down to cash

Martin states the rule flatly: "i slutändan finns det bara en enda anledning till att företag misslyckas och det är att man får slut på pengar" — in the end there's only one reason companies fail, and that's running out of money. ▶ 21:35 Everything else — strategy, product quality, market timing — is downstream of that constraint; raising capital, managing burn, and designing margin aren't peripheral tasks, they're the actual game.

That reframes bankruptcy itself. As an engineer, Martin expected the trigger to be objective — some line you cross. Instead he found it enormously subjective: "Så länge man har en plan för hur man ska rädda bolaget och tror på den planen, då kan man jobba vidare för att rädda ett bolag. När man inte längre kan tro på en plan... då är en konkurs framtvingad" (as long as you have a plan you believe in, you keep working to save the company; bankruptcy is forced only once no credible plan is left). ▶ 22:36 Bankruptcy, in other words, isn't a balance-sheet event — it's the moment a founder's own belief runs out.

Underneath the cash problem sat a design problem. Narrative priced for growth volume rather than sustainable cash flow — a strategy choice, not an accident — and it left some channels with roughly 10% margin against the 30–40% the business actually needed to survive. Martin says perfect execution couldn't have fixed that; the margin had to be designed in from the start, a lesson he later heard echoed by Instacart's Max Mallen, who now treats unit economics as the company's primary metric ahead of growth. ▶ 28:11

Narrative's collapse, told straight

His summary judgment on the whole arc: the worst thing that can happen to a company isn't failure — it's stagnation, chewing along year after year without getting anywhere. "Det värsta som kan hända med ett företag är egentligen inte att man misslyckas, utan att man inte kommer någon vart och tuggar på år efter år." Breakit ↗

Ship "ready enough," resist the apology feature

Two disciplines Martin credits with keeping Narrative Clip 1 sane, and that generalize well beyond it:

Decide when it's ready enough, on purpose. Good engineers need to be perfectionists — but left unmanaged, that trait produces a product that's never done. "It's really important to make the decision when you finally say 'it's ready' or 'it's ready enough.'" SlashGear ↗ Shipping with known gaps — Narrative Clip 1 launched with search removed because it "doesn't yet work properly" — teaches you what users actually need faster than polishing in isolation ever could. SlashGear ↗

Don't apologize for a delay by adding scope. Piling new promises onto a slipping deadline is a doom loop — you can end up never being done. The discipline is to ship the original commitment on schedule and let real usage tell you what to prioritize next. SlashGear ↗

Own the problem before you assign blame

A recurring failure mode Martin returns to in his Co-creating with AI conversations with Rasmus Adler Wahlberg is collaborative work breaking down over unclear ownership — and the fix is always ownership, not blame.

The concrete case is a Visa development team that took on building their own CMS with breezy overconfidence — "the development team sort of had the motto of how hard can it be?" — and ended up badly overcomplicating and overrunning it. ▶ 4:50 Rasmus's diagnosis: co-creation needs genuinely clear boundaries around ownership — defined timeboxes, defined budget — and their absence, not the team's competence, was the real cause. ▶ 6:05

Martin and Rasmus generalize this into a principle: when co-creation fails because boundaries were unclear, the leader must own that failure and re-establish boundaries — not unilaterally seize back control. "If you've failed to establish boundaries, then you need to own that and re-establish boundaries so that it doesn't become taking back ownership, it becomes clarifying ownership." ▶ 20:39 The framing draws explicitly on Jocko Willink's Extreme Ownership and the Navy SEAL culture of taking responsibility rather than pointing fingers ▶ 23:45 — because, as Martin puts it, owning the problem is simply "a more positive starting point" than blame, even when responsibility is genuinely shared. ▶ 24:56

That same conversation extends into a broader view of leadership under pressure: it should be distributed and fluid, earned through competence rather than hierarchy — "almost like fluid hierarchy," where you lead in one context and follow in the next, modeled on organizations like Burning Man and Zappos that widen access to leadership without flattening it into false equality. ▶ 21:15

What carries a founder through the dips

Losing both parents to cancer as a young adult is the formative event Martin traces his urgency back to — "When you lose your parents, you realise that you don't live forever... I can't wait until later to fulfil my dreams, I have to live my dream now." European CEO ↗ · Phys.org ↗ He names belief and passion, in the same register, as non-negotiable: "det är oerhört viktigt att tro på det man gör" (it's enormously important to believe in what you're doing) — because without personal passion, founders simply cannot get through the difficulties they inevitably meet. ▶ 12:46 Years later he puts the same idea in founder-roller-coaster terms: passion is what carries you through the valleys where things aren't fun, toward the next high. ▶ 33:02 And in the same breath, he flags people as the real variable — "the people you get to hang out with... choose them wisely," since a startup journey means spending years with your co-founders. ▶ 33:24

Two more recent data points show the same conviction still operating, in opposite directions. In 2017 he killed a fintech venture before launch rather than wait on uncertain regulatory approval — "Vi ville kunna gasa på direkt... Då insåg vi att det var bättre att inte dra igång alls" (we wanted to be able to move fast right away; we realized it was better not to start at all) — and still called the six months of work "not wasted time in any way... we learned an enormous amount and had a lot of fun besides." Breakit ↗ By contrast, in 2024 he pivoted away from a voice-AI project roughly two months from beta — and calls it a decision he "really regret[s]," admitting he chased San Francisco hype instead of staying home to finish the code. ▶ 14:19 ▶ 14:45 The two moments read as a matched pair: killing a project cleanly, before sunk cost sets in, is fine; abandoning one that's nearly done is the mistake he still carries.

Underneath it all sits his stated mantra, and a thesis he elaborates at length in a talk about wearable cameras: strive to create value, not success. Webit ↗ "If you manage to create success without value, what have you really accomplished? But if you create value but are not very successful, then you still have created something." ▶ 19:46

Resilience for the AI era

Martin extends the same failure-and-recovery logic forward to AI-driven trust erosion. As deepfakes and fraud (he cites a $25M deepfake Zoom fraud and a spoof-email attack aimed at Narrative itself) make impersonation cheap, both organizations and society need to be built with resilience and antifragility in mind — concepts he expects to move to the forefront specifically because of AI. ▶ 25:35 The practical response, in his view, is layered trust infrastructure: better technical authentication (2FA, digital signatures, cryptographic verification) paired with new social protocols — flatter organizations, tighter supply-chain verification — since trust, once eroded, raises the friction cost of doing any business at all. ▶ 22:19

He also applies a failure lens to AI disruption itself. On professional services, he notes that an AI-native competitor — a translation agency that adopts AI internally — has a choice once it has the efficiency gain: quietly pocket the margin at old prices, or compete openly on price and force the market to reset. ▶ 4:37 And on autonomous AI specifically, he argues the industry got lucky with sequencing: had ChatGPT launched with AutoGPT's autonomy from day one, the public backlash would have been immediate and severe — "really super black headlines about [AI] coming to steal our jobs" — rather than the gentler, gradual rollout that actually happened. ▶ 5:10 He goes further, wondering aloud whether some AI "doomers" spread fear deliberately to preserve a competitive advantage they'd rather keep to themselves. ▶ 26:01

Worth remembering