The last 10% of software reliability can cost as much as the rest of the build
MoonL88537 · x · 2026-07-23
The post argues that getting software to 85%–90% is usually enough to ship something useful, but the final 10% of reliability can cost as much as the rest of the product effort combined.
The conclusion is blunt: for many products, truly reliable software is economically irrational unless the stakes are high enough to justify the extra work. The quoted thread frames this as a broader critique of how vulnerable modern software is to exploits.
More from AGI Musings
- AI capabilities are improving faster than the way we measure them — ravi_iitm · 2026-07-23
- Deb Raji says frontier AI governance still misses state and municipal impacts — rajiinio · 2026-07-23
- Raji says FAccT still shaped AI policy more than any other CS conference — rajiinio · 2026-07-23
- JacquesThibs on AGI: You Can't Counterexample Your Way Through the Singularity — JacquesThibs · 2026-07-23
- AI governance needs independent oversight now, author says, with an IVO model ready to launch — ghadfield · 2026-07-23
- AI progress is now hitting physical-world constraints, not funding bottlenecks — finbarrtimbers · 2026-07-23