Blog post: https://lucumr.pocoo.org/2026/3/20/some-things-just-take-time/
Author: Armin Ronacher. Published March 20, 2026.
Ronacher's central argument is that in the current AI-accelerated era, the tech industry is dangerously undervaluing the irreplaceable role of time itself — not clock time or compute time, but the slow, compounding kind that builds trust, quality, community, and maturity. He uses the metaphor of a tree: a 50-year-old oak cannot be purchased or fast-tracked, only grown. He contends that the same is true for durable software, open-source communities, and company-customer relationships.
1. The Tree Metaphor — Time as an irreplaceable input. Ronacher opens with the observation that old trees, old gardens, and old buildings carry value precisely because of their age. He draws a parallel to how society already acknowledges this: we set age minimums for important activities, and we pay premiums for things with embedded time (aged properties, craftsmanship). This frames his argument not as nostalgia, but as an appeal to something we already intuitively understand.
2. "Friction Is Good" — Processes exist for a reason. He pushes back against the industry reflex to eliminate all friction. Compliance processes like SOC2, cooling-off periods for major decisions, and code reviews aren't obstacles — they are the point. The desire to automate away every human checkpoint with AI-driven shortcuts risks removing the very mechanisms that produce quality and accountability. Doing something right once doesn't prove you can sustain it.
3. "Vibe Slop at Inference Speeds" — Speed without direction produces disposable software. This is the sharpest section. Ronacher observes that AI's ability to generate code quickly is creating pressure to strip away reviews, design processes, and infrastructure configuration — everything that "slows the pipeline." The result: software with a shelf life of months, not decades. He points to YC startups that simply vanish without notifying customers, and open-source projects with a week of commits before abandonment. He frames this not as healthy iteration but as a breakdown of the trust required for lasting relationships.
4. "My Time" — The productivity paradox. Perhaps the most personally candid section. Ronacher admits that despite being deeply embedded in AI tooling, he has less time, not more. He identifies a competitive trap: any time saved is immediately consumed by additional output, and anyone who pauses is outmaneuvered by someone who doesn't. The promised dividend of AI-driven efficiency is being entirely captured by competitive escalation, not reinvested in quality or rest.
5. The Return to Trees — Commitment as the real product. He closes by reflecting on his own nearly two decades of open-source maintenance and 10 years at a single startup (Sentry). He attributes their success not to discipline or virtue, but to the simple act of showing up repeatedly until the project developed roots deeper than any individual day's enthusiasm. The essay ends with a personal note: he recently planted a tree with his son Colin, and he's not in a rush.
The velocity-vs.-direction distinction. The top HN commenter (Chris_Newton) crystallized this beautifully: velocity is a vector quantity — speed only helps if you're heading the right way. AI coding tools excel at prototyping and exploration (plotting a course) but often flounder on implementation for existing systems because they lack deep contextual understanding. Speed without direction is waste.
The Jevons Paradox of developer time. Ronacher's observation about time savings being immediately consumed maps directly to the economic concept of Jevons Paradox — making a resource more efficient to use can increase total consumption rather than decreasing it. Multiple commenters confirmed this from personal experience, with one FAANG engineer describing how leadership simply ratcheted up productivity expectations to absorb all AI-enabled gains, producing burnout rather than better work.
Trust is a slow variable. The argument that durable open-source projects, customer relationships, and communities cannot be manufactured on short timescales is fundamentally about trust as a slow variable. Trust accumulates through repeated demonstration over time. This is what Ronacher means when he says commitment can shelter and grow other people — it becomes infrastructure that others can rely on.
The friction-as-feature insight. This challenges a deep assumption in engineering culture. Compliance, review processes, and even bureaucratic slowness often serve as forcing functions for reflection, error-catching, and institutional learning. Removing them optimizes for throughput at the cost of reliability and wisdom.
Pushback on the luxury-goods analogy. Commenter "titanomachy" pointed out that people pay premiums for Swiss watches and Hermès bags because they are status symbols, not because of the time embedded in crafting them. A grandmother's hand-knitted sweater takes just as long but has near-zero market value. This is a fair critique — Ronacher conflates time-as-quality with time-as-scarcity/brand, and the argument would be stronger if it focused purely on the compounding effects of sustained commitment rather than luxury goods.
AI as a sparring partner, not a replacement. Several commenters (ChrisMarshallNY, bytefish) described a middle path: using LLMs interactively for debugging, validating API designs, and exploring large codebases — while maintaining human judgment about direction and quality. This nuances the article's somewhat binary framing. The issue isn't AI itself but the ideology of total automation that sometimes accompanies it.
The Theory of Constraints applies. Commenter "alexpotato" (building a game with AI assistance) noted that even with AI-accelerated code generation, playtesting and design decisions still occur at human speed. The bottleneck simply shifts. This is a powerful complement to Ronacher's argument: AI doesn't eliminate time constraints, it relocates them to wherever human judgment, taste, and experience remain necessary.
The FAANG burnout angle. The comment from a FAANG employee describing how AI productivity gains are being absorbed entirely by raised expectations and internal competition adds an important structural dimension. The problem isn't just cultural (hustle mindset) — it's systemic. When productivity tools are deployed in competitive organizations, the gains flow to the employer, not the employee, and the result can be worse working conditions despite better tools.
Counterpoint: some friction is genuinely wasteful. Not all commenters agreed with the "friction is good" thesis. The implicit counter-argument is that there's a difference between meaningful friction (code review, design thinking, cooling-off periods) and bureaucratic friction (pointless compliance theater, cargo-cult processes). Ronacher's argument is strongest when applied to the former and less compelling as a blanket defense of all slowness.
The post hit 767 points because it articulates an anxiety that many in the tech community feel but struggle to name: the sense that the current AI acceleration is producing speed without wisdom, output without durability, and busyness without meaning. Ronacher's credibility as someone who has actually maintained open-source projects for nearly 20 years and spent a decade at a single company gives the argument weight that it wouldn't carry from a newer voice. The tree metaphor is simple and sticky, and the personal vulnerability about feeling time-poor despite being at the center of the AI boom gives it emotional resonance. This is, at its core, a meditation on what gets lost when an industry optimizes exclusively for speed — and the answer, Ronacher suggests, is everything that makes the work worth doing in the first place.