One Week Later, And I’m Even Less of a Coder
The Hockey Stick Hit in Early 2026
A week ago, I hit publish on “I Don’t Write Code Anymore.” I called myself a Maestro, waving hands while agents built service meshes and refactored architectures as I ate pizza. I felt like a fraud, prompt orchestration engineer, syntax sommelier, product manager for AGI. Whatever that means.
I thought I’d captured the vibe: existential uncertainty mixed with superpowers.
Then 2026 started. And holy hell, the hockey stick arrived early.
The Acceleration Feels Different Now
Last week, I delegated features to subagents and reviewed diffs while sipping coffee. Solid 5-10x velocity. Impressive.
This week? I barely open the IDE. Late-2025 drops like Claude Opus 4.5 handle entire multi-day tasks in hours. I describe a vague idea: ”build a resilient event-sourcing backbone with CQRS, add OpenTelemetry tracing, make it idempotent”, and come back to a polished implementation, tests, and migration plan.
I didn’t write a single line. Not even the prompts were that detailed. The agent asked clarifying questions, iterated on its own, and fixed its mistakes.
I’m not a Maestro anymore. I’m... the guy who says “go” and hopes it doesn’t burn the orchestra down.
It’s Not Just Me
I thought I was deep in the bubble. Then I looked around.
Claude Code Creator Boris Cherny: “The last month was my first as an engineer that I didn’t open an IDE at all. Opus 4.5 wrote around 200 PRs, every single line.”
These aren’t hype merchants. These are the people shipping the future, feeling the exact same whiplash.
Jaana’s Wake-Up Call
A Principal Engineer at Google working on the Gemini API admitted her team spent a year building distributed agent orchestrators. Slow progress, debates, no alignment.
She gave Claude Code a brief description of the problem. It generated a functional version in one hour.
She wrote: “I’m not joking and this isn’t funny.”
That post exploded because it was raw truth from inside Big Tech. And remember, that was before the absolute latest Opus 4.5 tweaks. Today? It’d probably do it in minutes.
The Paradox Persists
Productivity is exploding, but it doesn’t always feel like winning.
Some days, agents ship magic. Other days, I spend hours reviewing hallucinations, untangling weird architectural choices, or debugging subtle bugs that human intuition would have caught immediately.
The review bottleneck is real. Generation is near-instant. Verification? Still human-speed.
And the deeper question keeps me up: what happens to muscle memory? If I never touch the code, do I still deeply understand the system? Or am I slowly becoming the manager who “used to code”?
What 2026 Might Look Like
Three rough paths ahead:
1. Acceleration (feels most likely): Agents get reliable enough that solo builders ship what used to take teams. One-person unicorns become normal. The gap widens between orchestrators and those resisting.
2. Plateau: Quality debt, security holes, and trust issues force a slowdown. Enterprises clamp down. Infinite velocity without guardrails creates expensive messes.
3. Equilibrium: We nail the human-AI loop. Humans focus on intent, architecture, taste. Agents handle execution. Roles evolve to “Agent Orchestrator” or “Intent Designer.” The craft changes, but doesn’t die.
I’m betting on a mix of 1 and 3. The tools are too powerful to ignore, but we’re not dumb enough to trust them blindly forever.
Final Thought
A week ago, I thought I’d transcended coding.
Now I realize I was just at the starting line.
The job isn’t gone. It’s mutated. Again. Faster than any of us expected.
If you’re reading this and feeling behind—join the club. Karpathy, Cherny, Osmani, Dogan... we’re all there.
The only way forward: build something this week with the latest agents. Ship something small. Feel the speed. Feel the unease.
Then tell me what you call yourself now.
Because “Maestro” definitely isn’t it anymore.
What’s your new role in 2026? Drop it in the comments—I’m collecting suggestions for my next identity crisis.