Architects already solved non-determinism with eventual consistency
Architects reject AI-generated code because it is non-deterministic. They solved that exact problem a decade ago in their event-driven systems, and the saga pattern shows where the correctness comes from.
Written by
-
Benedikt Stemmildt Founder & Co-CEO
Published: 17 September 2026 Updated: 1 min read
The most common objection from senior architects is the same sentence, almost word for word. The agent is non-deterministic. Run it twice, get two different results. You cannot build on that.
It sounds like a reason to wait. It is actually a problem they already solved.
They gave up strong consistency once before
Ten years ago these same architects stopped insisting their distributed systems be strongly consistent. Two services, one source of truth, every read perfectly in sync, every write blocking until everyone agreed. That model does not scale, and they knew it. So they let go of it. They accepted that the system passes through inconsistent states and converges to a correct one. Eventual consistency. They built their entire event-driven architecture on it.
A non-deterministic generation step has the same shape. The agent does not produce one guaranteed-correct output. It produces a candidate that has to converge to correct.
The constraint is the convergence layer
So strong consistency for code generation is solved by giving it up, the way it was solved for data. The constraint is now the convergence layer: the thing that pulls the candidate to correct.
Architects already built that layer too. When a distributed transaction cannot be atomic, you do not block. You add a saga: each step has a compensating action that fixes the state if a later step fails. Non-determinism upstream, correctness restored by a downstream gate.
Review critics are compensating actions
That is exactly what review critics are. The PR reviewer, the post-merge verifier, the test suite the diff has to pass. They are compensating actions for a generation step that does not promise determinism. Not a workaround, but the same correctness model applied one layer up.
The architects who reject the agent and the architects who built the saga pattern are the same people. They are holding the answer.
What would your team build differently if it treated agent output as eventually consistent, not unreliable?
Further reading
Written by
-
Benedikt Stemmildt · Founder & Co-CEO
More on Specification and review
- Fast agents push the bottleneck back to the requirements Delivery was built around slow coding. With code, review and deployment fast, questions return faster than anyone answers them, and the constraint moves to where requirements are written.
- YAGNI still holds, for a different reason Cheap code tempts teams to generate the flexible version just in case. But writing was never the only cost. Reading, debugging, and changing the code still cost a human, and YAGNI now protects those.
- Agents stall on unclear requirements, and a spec file fixes it A pilot team blamed the agent until it looked at what the agent was handed. The fix was one spec file in the repo, with QA's test cases written before any code.
Test the constraint in your system
The Factory Readiness Assessment connects this research to your value stream, your data, and the next decision.
Training on this topic: Agentic Engineering for Developers and QAWant to check one team first? The self-check: twelve yes or no questions, three minutes, instant result, no account. Take the self-check