Key Takeaways
- AI agents eliminate excuses for product backlogs. Anker Goyel states, “there's just no excuse to not have rigor... we don't really have a backlog,” meaning every "paper cut" or performance issue gets addressed immediately.
- Product building with AI is now “carving rather than constructing.” AI rapidly generates too many features, so the primary engineering task shifts to removing complexity to improve user experience.
- Continuous Integration (CI) becomes a strategic investment for velocity. Instead of shipping poor quality code, Goyel's team pauses to “improve CI so that we earn the ability to move faster.”
- For AI products, the engineering team's number one job is to build a robust feedback loop. This pipeline must “summon from the ether of real world data and turn that into eval.”
- The development cycle becomes a constant loop of refining evals: “Close the session... and then I improve the evals and then I try from scratch again.”
The Method: "No Excuse" for Quality
For Anker Goyel, co-founder of How I AI, the rise of AI agents means one thing for engineering teams: there's "no excuse" for low quality or backlogs. Forget the old way of prioritizing a laundry list of bugs and missing features. In an AI-powered world, if someone complains about a UI "paper cut" or a performance hiccup, the default is to fix it now. This isn't just about speed; it's about a philosophical shift where rigor and performance become non-negotiable defaults.
Goyel describes product building with AI as a complete inversion of the traditional process. “Product building and code writing is now looks like carving rather than constructing,” he explains. AI agents can churn out features, buttons, and code at breakneck speed. The real work, the hard work, comes in taking that initial, often feature-bloated output and meticulously removing complexity. It’s a sculptor’s approach: start with too much material and pare it down until only the essential, elegant form remains.
This high-velocity, high-quality approach hinges on a significant investment in Continuous Integration. Goyel's team spends far more time on CI than they used to, viewing it not as a cost center but as a prerequisite for speed. “If we are really good at CI then we're able to move faster,” Goyel says. If velocity drops, they don't push out "crappy stuff." Instead, they pause, shore up their CI, and "earn the ability to move faster." It's a pragmatic, almost mercenary view of CI: optimize your tools before you expect peak output.
Crucially for AI products, the engineering team's most important task becomes building an unbreakable feedback loop. This means creating a pipeline that can automatically pull real-world data and user interactions from "the ether" and transform them into actionable evaluations, or evals. The goal is a continuous cycle of learning and improvement. Goyel’s process is stark: a session runs, then he stops, "improve the evals and then I try from scratch again." It's an engineering mindset applied directly to product iteration, ensuring every bug or sub-optimal interaction feeds directly back into the system for a fresh, improved attempt.
Where This Breaks Down
Goyel’s "no excuse" philosophy demands significant upfront investment. Building a world-class CI pipeline and a sophisticated, automated feedback loop for AI evals isn't cheap or easy. A very early-stage startup with limited engineering resources might find this difficult to implement from day one. They might initially prioritize shipping something over perfect, backlog-free quality, even if it accumulates a few "paper cuts."
This method also assumes a certain maturity of AI agents. If your agents are unreliable, or if the "too many features" problem means outright broken or unusable code, then "carving" becomes less about refinement and more about triage. It also requires a cultural shift: a team must be fully bought into radical transparency on quality and a constant, almost surgical approach to product iteration. Without that buy-in, the system might lead to burnout rather than improved quality.
What to Do With This
Take a hard look at your product's "paper cuts." Pick one small, irritating bug or UI friction point that's been lingering in your backlog. This week, commit to fixing it immediately. No excuses, no moving it to the next sprint. Experience what it feels like to eradicate a backlog item instantly.
Next, audit your Continuous Integration setup. Is your team truly investing in CI as a speed multiplier, or is it just a necessary evil? Identify one bottleneck in your CI pipeline that, if improved, would demonstrably accelerate your development or testing cycles. Prioritize fixing that bottleneck this month, even if it means temporarily pulling engineers off feature work.
Finally, if you're building an AI product, sketch out a minimal viable feedback loop. How can you take one simple, real-world user interaction—say, a user editing an AI-generated output—and automatically feed that data back into an eval system? Start building the simplest pipeline to achieve this by the end of the quarter. Your goal is to get some real-world data fueling your evals, not a perfect system from the start.