68 posts tagged with "testing"

Most teams ship an LLM feature, then spend weeks arguing about whether it's actually good. The evaluation question gets deferred because building a labeled dataset feels like a separate project. By the time you have ground truth, you've also accumulated two months of silent regressions you can never diagnose. This is backwards. You can get a meaningful quality signal in week one — before a single annotation is complete — if you know which techniques to reach for and where each one breaks.

This post is a field guide to annotation-free evaluation: the reference-free methods that work, the conditions they require, and the specific failure modes that will fool you if you're not careful.

The 12-Factor App doctrine made dev/prod parity famous: keep development, staging, and production as similar as possible. For traditional web services, this is mostly achievable. For LLM applications, it is structurally impossible — and the gap is far larger than most teams realize.

The problem is not that developers are careless. It is that LLM applications depend on a class of infrastructure (cached computation, living model weights, evolving vector indexes, and stochastic generation) where the differences between staging and production are not merely inconvenient but categorically different in kind. A staging environment that looks correct will lie to you in at least seven specific ways.

A medical AI deployed to a hospital achieves 97% accuracy in testing. It passes every internal review, gets shipped, and then quietly fails to detect parasitic infections when parasite density drops below 1% of cells — the exact scenario where early intervention matters most. Nobody notices until a physician flags an unusual miss rate on a specific patient population.

This is the long-tail coverage problem. Your aggregate metrics look fine. Your system is broken for the inputs that matter.

Most teams ship LLM changes the way they shipped web changes in 2005 — they run some offline evals, convince themselves the numbers look fine, and push. The surprise comes on Monday morning when a system prompt tweak that passed every benchmark silently breaks the 40% of user queries that weren't in the eval set.

Shadow traffic is the fix. The idea is simple: run your candidate model or prompt in parallel with production, feed it every real request, compare the outputs, and only expose users to the current version. Zero user exposure, real production data, and statistical confidence before anyone sees the change. But applying this to LLMs requires rethinking almost every piece of the implementation — because language models are non-deterministic, expensive to evaluate, and produce outputs that can't be compared with a simple diff.

Most teams building LLM applications discover the same problem around week three: every time someone runs the test suite, it fires live API calls, costs real money, takes 30+ seconds, and returns different results on each run. The "just hit the API" approach that felt fine during the prototype phase becomes a serious tax on iteration speed — and a meaningful line item on the bill. One engineering team audited their monthly API spend and found $1,240 out of $2,847 (43%) was pure waste from development and test traffic hitting live endpoints unnecessarily.

The solution is not to stop testing. It is to build the right kind of development loop from the start — one where the fast path is cheap and deterministic, and the slow path (real API calls) is reserved for when it actually matters.

When Anthropic deprecated a Claude model last year, a company noticed — but only because a downstream parser started throwing errors in production. The culprit? The new model occasionally wrapped its JSON responses in markdown code blocks. The old model never did. Nobody had documented that assumption. Nobody had tested for it. The fix took an afternoon; the diagnosis took three days.

That pattern — silent behavioral dependency breaking loudly in production — is the defining failure mode of model migrations. You update a model ID, run a quick sanity check, and ship. Six weeks later, something subtle is wrong. Your JSON parsing is 0.6% more likely to fail. Your refusal rate on edge cases doubled. Your structured extraction misses a field it used to reliably populate. The diff isn't in the code — it's in the model's behavior, and you never wrote a contract for it.

With major providers now running on 60–180 day deprecation windows, and the pace of model releases accelerating, this is no longer a theoretical concern. It's a recurring operational challenge. Here's how to get ahead of it.

Ask any AI engineering team if they test their prompts and they'll say yes. Ask if a bad prompt can fail a pull request and block a merge, and you'll get a much quieter room. The honest answer for most teams is no — they have eval notebooks they run occasionally, maybe a shared Notion doc of known prompt quirks, and a vague sense that things are worse than they used to be. That is not testing. That is hoping.

The gap exists because prompt testing feels qualitatively different from unit testing. Code either behaves correctly or it doesn't. Prompts produce outputs on a spectrum, outputs are non-deterministic, and running enough examples to feel confident costs real money. Those are real constraints. None of them are insurmountable. Teams that have built prompt CI that actually blocks merges are not spending fifty dollars a build — they're running in under three minutes at under a dollar using a few design decisions that make the problem tractable.

Shipping a feature that calls an LLM is easy. Knowing whether the next version of that feature is better or worse than the one in production is hard. Traditional CI/CD gives you a pass/fail signal on deterministic behavior: either the function returns the right value or it doesn't. But when the function wraps a language model, the output is probabilistic — the same input produces different outputs across runs, across model versions, and across days.

Most teams respond to this by skipping the problem. They run their unit tests, do a quick manual check on a few prompts, and ship. That works until it doesn't — until a model provider silently updates the underlying weights, or a prompt change that looked fine in isolation shifts the output distribution in ways that only become obvious in production at 3 AM.

The better answer isn't to pretend LLM outputs are deterministic. It's to build CI gates that operate on distributions, thresholds, and rubrics rather than exact matches.

Your content moderation service returns {"severity": "MEDIUM", "confidence": 0.85}. The downstream billing system parses severity as an enum with values ["low", "medium", "high"]. A model update causes the service to occasionally return "Medium" with a capital M. No deployment happened. No schema changed. The integration breaks in production, and nobody catches it for six days because the HTTP status codes are all 200.

This is the foundational problem with API contracts for LLM-backed services: the surface looks like a REST API, but the behavior underneath is probabilistic. Standard contract tooling assumes determinism. When that assumption breaks, it breaks silently.

Thirty percent of generative AI projects are abandoned after proof of concept. Ninety-five percent of enterprise pilots deliver zero measurable business impact. Gartner projects 40% of agentic AI projects will be canceled before the end of 2027. These aren't failures of the underlying technology — they're failures of the gap between demo and production.

The demo-to-production failure pattern is predictable, repeatable, and almost entirely preventable. It happens because the conditions that make a demo look great are systematically different from the conditions that make production work. Teams optimize for the former and get ambushed by the latter.

Your eval suite says 94% accuracy. Users report the feature is broken for names that aren't "John" or "Alice." Both things are true, and the gap between them has a name: your curated test set encodes only the failure modes you already imagined.

Property-based testing (PBT) was invented in 1999 to expose exactly this class of blind spot in deterministic software. Applied to LLM outputs, it generates tens of thousands of adversarial input variants automatically, probing domain boundaries that hand-written test cases structurally cannot reach. A 2025 OOPSLA study found that on average each property-based test discovers approximately 50 times as many mutant bugs as the average unit test. A separate study measured that PBT and example-based testing (EBT) fail on different bugs — combining both raised detection rates from 68.75% to 81.25%. That 12.5-point gap is not rounding error; it represents an entire class of failure invisible to one approach.

This article is for engineers who already have eval suites and want to find the bugs those suites structurally cannot find.

Your customer service agent has been running reliably for three months. A routine model update rolls in on a Tuesday. By Wednesday afternoon, three downstream services are silently parsing the wrong fields from the agent's responses—the JSON keys shifted subtly but nothing returned an error. By Thursday you've traced a drop in order completions to a JSON field renamed from "status" to "current_state". The model updated, the agent stayed at v2.1.0, and nobody got paged.

This is the versioning gap that nobody in traditional API design had to solve. Semver works when you can deterministically reproduce outputs from a specification. AI agents can't make that promise. Yet downstream services depend on their behavior just as critically as they depend on any microservice API. The gap between "we tagged a release" and "downstream consumers are protected" has never been wider.