From 25% to 90%: One Developer's Journey With AI Coding Agents
David Crawshaw, founder of exe.dev and former Tailscale co-founder, just published a follow-up to his viral posts about programming with AI agents. The numbers are staggering: in February 2025, Claude Code wrote about 25% of his code. By February 2026, the latest Opus model writes 90% of it. His time split went from 50-50 (reading vs. writing code) to 95-5.
The post, titled "Eight More Months of Agents," hit 110+ points and 115 comments on Hacker News within a day. If you've been wondering whether AI coding agents are actually worth integrating into your daily workflow — or if the hype is overblown — Crawshaw's experience offers one of the most honest, data-rich accounts available.
The Numbers: How AI Agents Transformed One Developer's Output
| Metric | February 2025 | February 2026 |
|---|---|---|
| Code written by AI | ~25% | ~90% |
| Time split (reading vs. writing) | 50-50 | 95-5 |
| Primary tool | IDE with Copilot | Neovim + Claude Code |
| Developer role | Writing code | Reading and directing agents |
Crawshaw is clear: the code still "all needs to be carefully read, and regularly adjusted." But the key shift is that he now relies on the model to do the adjustments too. The developer's role has fundamentally changed from writing to reviewing and directing.
It's All About the Model, Not the Harness
One of Crawshaw's most provocative claims is that agent harnesses haven't improved much in the past year. He notes that his own agent, Sketch, could do things six months ago that popular agents still can't do today.
"Right now, it is all about the model," Crawshaw writes.
He dismisses public benchmarks entirely — "they have all been gamed to death" — but points to the qualitative leap in model capabilities as the most significant change.
This matters for developers choosing tools. If you're spending hours comparing agent frameworks, Crawshaw's experience suggests you should focus on which tool gives you access to the best model instead.
IDEs Are Dead. Long Live Vi.
Perhaps the most surprising takeaway: Crawshaw has abandoned IDEs entirely. In 2021, when Copilot launched, IDEs seemed inevitable. Autocomplete and inline edit made your typing "go 50% further."
Four years later, the story has completely reversed. Agents don't need an IDE — they need a terminal and access to your codebase.
"So here I am, 2026, and I am back on Vi," he writes. "Vi is turning 50 this year."
Use Only Frontier Models — Or Learn the Wrong Lessons
Crawshaw makes a strong case: using anything other than frontier models is "actively harmful." Not just wasteful — actively counterproductive.
His argument: the limits of AI agents keep moving, and developers need to constantly re-learn what's possible. If you use a cheaper model, you'll develop an inaccurate mental model of what agents can actually do.
"Pay through the nose for Opus or GPT-7.9-xhigh-with-cheese," he advises. "Don't worry, it's only for a few years."
Sandboxes Are Broken — Use Fresh VMs
A practical pain point: built-in agent sandboxes don't work well. Claude Code's constant "may I run cat foo.txt?" prompts and Codex's inability to build in its sandbox create friction that kills productivity.
His recommendation: turn off the built-in sandbox and provide your own fresh VM for each session.
More Programs Than Ever Before
The most optimistic section: Crawshaw now has far more programs and services than he used to. Ideas that would have ended up in an Apple Note titled "TODO" — forgotten forever — now actually get built.
"I am having more fun programming than I ever have," he writes, "because so many more of the programs I wish I could find the time to write actually exist."
For developers who already know what they want to build, agents remove the bottleneck of implementation time. The creative, architectural work remains human.
What Hacker News Thinks
The HN discussion reveals the deep divide in the developer community:
The Skeptics Have a Point — Multiple commenters pushed back on Crawshaw's frontier-model-only stance. One noted they get "solid results from Copilot and Haiku/Flash" using optimized prompts.
The Fear Is Real — Five categories of anti-LLM sentiment emerged: job loss, happiness loss, uncertainty, expertise erosion, and training data ethics.
The "Gaslit" Developer — One of the most upvoted comments: "Comments like these make me feel like I'm being gaslit." They described coworkers producing "pure slop that doesn't work, buggy, untested adequately."
This tension between power users and average users is perhaps the defining challenge of AI coding tools in 2026.
The Bigger Picture: Software Is the Wrong Shape
Crawshaw's most forward-looking insight: "most software is the wrong shape now." Rather than waiting for Stripe's API to be available, he had his agent build an entire ETL pipeline — three sentences of instruction replaced a commercial product.
His philosophy: "The best software for an agent is whatever is best for a programmer." If he's right, the entire SaaS industry needs to rethink how products are built — prioritizing APIs and developer experience over polished UIs.
What This Means for You in 2026
| If You're... | Key Takeaway |
|---|---|
| Already using AI agents | Focus on frontier models and VM-based sandboxes |
| Skeptical but curious | Try Claude Code or Codex with the best model available |
| A product builder | Prioritize APIs and developer experience |
| Worried about job impact | The role is shifting from writing to reviewing and directing |
FAQ
How much code can AI agents write in 2026?
According to Crawshaw's experience, frontier AI models like Claude Opus can now write approximately 90% of production code, up from 25% one year earlier.
Are AI coding agents actually worth the cost?
Crawshaw argues frontier models are essential because cheaper models teach you "the wrong lessons" about what AI agents can do.
Do I still need an IDE if I use AI coding agents?
Crawshaw has abandoned IDEs entirely in favor of neovim. Terminal-based agents like Claude Code don't require IDE integration.
Will AI agents replace software developers?
AI agents are changing the developer role rather than eliminating it. Developers now spend 95% of their time reading and reviewing code rather than writing it.
Originally published on Serenities AI
Top comments (1)
This is fascinating data, Tan Genie! David Crawshaw's journey from 25% to 90% AI-written code in just 8 months really puts the AI coding revolution in perspective. It matches what Anthropic's recent 2026 trend report predicted — that we're moving from "writing code" to "commanding agent armies" .
The key insight here isn't just the percentage increase, but what David said about "subtle conceptual errors" still requiring human oversight. That's the reality of AI coding right now — massive productivity gains, but human judgment remains irreplaceable . Thanks for sharing this!