Building got cheap. Keeping it running is the real job now
Most of the loneliness of a one-person product is not in building anymore. It is in the morning you open the dashboard and the thing you shipped in an afternoon is quietly on fire.
I wrote yesterday that code grades high at review and production disagrees. Bain just published the organizational version of the same argument, and their numbers are better than mine.
Transparency first: I read what Bain calls the reliability paradox — I have not rerun their survey. It sampled 293 tech and engineering leaders in April 2026 and dropped October 19. The headline: AI compresses the cost of building software but can inflate the cost of running it, unless operations get redesigned in parallel.
What does the reliability paradox measure?
A gap between two halves of the same system. Executives expect the most dramatic gains from AI in release speed and engineering cost efficiency — an almost 150% improvement over the next one or two years. The expected gain in operational resilience is the smallest of every metric they measured. Roughly half of organizations expect many crucial engineering tasks to be AI-led within two years, but only about 2% describe their engineering function as autonomous-first. Build velocity is compounding. Run capability is standing still — faster code aimed at an unchanged operating model.
Why did everybody invest in the half that produces demos?
Because the scoreboard lives there. Two-thirds of executives ranked applying AI to coding, testing, and reviewing as the most important productivity lever, while observability and incident automation were cited by roughly one in five. The run side of the house — monitoring, incident response, keeping the lights on — is unglamorous, so it got the smallest share of attention. The same instinct shows up outside enterprises: NN/G's October 2026 survey found the top reason UX practitioners avoid AI is output quality, with 54% citing concerns about the work coming out of it. Checking AI output takes real time, so everybody optimizes the generating and budgets almost nothing for the verifying.
What does this cost a one-person company?
You pay the run cost in the same person who was supposed to be building the next feature. AI lets you point a bigger stream of change at a product held up by one person's spare attention, and if an agent writes it, nobody wrote down how the system works in the first place. My opinion, and it is a strong one: a solo builder should take the Netflix example in that Bain brief the right way. Netflix routes 56% of memory-configuration errors to automatic fixes because it wired ownership, dependencies, and incident history into one platform. The lesson is not "buy tooling." The lesson is context has to exist before autonomy does. I do not want agents remediating my product; I want a product simple enough that I do not need remediation.
What to fix this week
Three moves, all free.
Shrink before you scale. Every unused service and third-party tool is a future morning of debugging. The ideal number of moving parts is the smallest that still ships.
Get told when it breaks. Error tracking with a short alert path is still the cheapest team member you will ever hire. If production is the only review that counts, it should be allowed to interrupt you.
Write the runbook down. Google's SRE tooling makes agents fetch an approved playbook through a function call instead of improvising. You are not that far from it: if the fix for every failure you know about lives in a doc, you are one agent away from the same setup. If it lives in your head, you are always the single point of failure.
Building got cheap. The bill shows up in running, and the founders who write down how their systems work get to keep both the speed and the sleep.
Frequently asked questions
AI compresses the cost of building software while inflating the cost of running it, unless operations get redesigned in parallel — the pattern Bain named in its October 2026 brief. In a survey of 293 tech and engineering leaders, expected gains in operational resilience trailed every other metric measured, while executives expected almost 150% improvement in release speed and cost efficiency.
About the author
mosh
mosh is a product designer for growth, working with design thinking and ever-improving design systems. What matters: fixing conversion, whether in B2B dashboards or direct-consumer apps.
Keep reading
- Escape Ludo: game design
A covid quarantine project. Sprite design, character formation, levels, and inventory UI for a multiplayer Ludo game.
- Go Laundry: service booking platform
One platform to book cleaning services in Qatar. Art-directed brand mascots, app UI, and UX research.
- SaaS activity tracking: designing an admin log teams trust
Admins needed to see every team action in one place. Research with two managers and eight user stories turned a confusing log into filters, search, share, and export.