AI agents haven't accelerated the traditional SDLC — they've dissolved it. Sequential phases collapse into a tight loop: Intent + Context → Agent builds → Observe → Repeat. What remains: Context Engineering and Observability.
AI agents don’t make the SDLC faster — they eliminate sequential handoffs. Requirements, Design, Implementation, Testing, Review, Deployment: all phases collapse into a simultaneous generation-and-verification cycle. Tane contrasts the old flow (Requirements → Design → Code → Test → Review → Deploy → Monitor) with the new one: Intent + Context → Agent → Build + Test + Deploy → Observe → Loop.
The engineer’s role shifts from writing code to providing context and direction. Tane’s key statement: “The SDLC is dead. The new skill is context engineering. The new safety net is observability.” The quality of agent-driven development depends entirely on context quality — not on process or ceremony.
Sprint Planning, Estimation, Pull-Request Reviews — all process rituals that become obstacles in agent-driven workflows. Tane explicitly calls the PR queue a “Fake Bottleneck” that only exists because human rituals are forced onto machine workflows. When agents generate hundreds of PRs daily, human review becomes a bottleneck rather than quality control.
Monitoring is the only phase that survives — but must fundamentally transform. Traditional dashboards for human interpretation aren’t sufficient when agents deploy hundreds of changes daily. The observability layer becomes the feedback mechanism driving the entire loop — not a phase at the end, but the connective tissue of the whole system.
Engineers who started their careers after the launch of Cursor don’t know Sprint Planning or multi-day Pull Request Reviews at all. They “just build things” — directly from description to shipped feature. Tane sees this not as a deficit, but as empirical validation of his thesis.
01 Context Engineering as a Design Competency: If context becomes the central resource — isn’t that exactly what good designers and product people have always done? Bringing user context, business context, and technical constraints together? How do we position this as a Lab?
02 Where Is Product Design? Tane only addresses architecture. What happens with User Research, UX, Service Design in an agent-driven world? Do these phases also collapse — or do they become more important?
03 Govtech Reality Check: Our Govtech projects have documentation requirements, accessibility mandates, auditability. How would we adapt the “Tight Loop” without violating regulatory requirements?
04 Fake Bottleneck or Real Quality Assurance? Is the PR queue really just a ritual — or are there contexts where human reviews serve a function that agents can’t (yet) cover?
05 Observability as a Business Model: If observability becomes the connective tissue, is there an opportunity here? Can we think of “Design Observability” as a service — the ability to measure whether a product does what it’s supposed to?
Context Engineering The competency of providing the right context to an AI agent — instead of writing code yourself. Encompasses formulating intent, structuring requirements, and curating relevant information for the agent.
Observability The ability to understand a system’s behavior from its outputs. In agent-driven workflows, observability transforms from a monitoring dashboard into the central feedback mechanism driving the entire build-observe cycle.
Feature Flag A mechanism that allows features to be toggled on or off in production — without redeployment. Enables progressive rollouts and automatic rollbacks when issues arise.
Tight Loop A compressed development cycle where intent, build, test, deploy, and observe happen near-simultaneously. Replaces the sequential SDLC with its separate phases and handoffs.