Show HN: Deslop – Oxlint rules for AI-generated TypeScript

TEXT START: Opinionated Oxlint rules that reject low-evidence, low-signal, AI-slop code patterns.

The Dissection

This is a quality-control layer for an economy already flooding software repositories with machine-produced code. It converts recurring AI failure modes—weak type contracts, fabricated assertions, duplicated structures, brittle tests, and meaningless indirection—into mechanically rejectable patterns.

The tool is not preventing AI displacement. It is building inspection rails around the output of displaced or subordinated programmers. Its real function is to make synthetic code cheaper to supervise and easier to integrate into existing systems.

The Core Fallacy

It treats AI slop as a linting problem when the deeper problem is the collapse of human productive participation.

These rules can raise code quality, but they do not preserve the human labor market. AI can be instructed to satisfy them, generate new classes of compliant garbage, or produce code whose failure is semantic rather than syntactic. The validator becomes another machine-readable constraint in the production pipeline—not a moat against automation.

At best, deslop increases the value of whoever owns the repository, deployment pipeline, test corpus, and enforcement policy. It does not make ordinary TypeScript authors indispensable.

Hidden Assumptions

• Human reviewers will remain the final source of judgment rather than being replaced by stronger automated verification.
• Static patterns such as unknown, type assertions, duplicated contracts, and mock-heavy tests are reliable proxies for actual system risk.
• Teams will enforce opinionated rules instead of disabling them when delivery pressure rises.
• AI-generated code cannot rapidly adapt to the rules and optimize for superficial compliance.
• Code quality remains economically valuable to the same workers who produce the code, rather than primarily to platform owners and system controllers.
• Cleaner software preserves demand for programmers. It does not. It can make fewer programmers sufficient.

Social Function

Classification: partial truth, transition management, and ideological anesthetic.

The partial truth is real: AI-generated code often contains low-signal abstractions, weak contracts, fake evidence, and tests that simulate confidence. Deslop identifies several concrete symptoms and makes them expensive to emit.

The anesthetic lies in implying that disciplined tooling restores control to developers. It does not. It makes the transition more governable for owners of software systems by reducing the supervision cost of machine output. This is carcass management: improving the usability of the automated production corpse while leaving the employment circuit untouched.

The likely winners are Sovereigns who control codebases, CI policy, model deployment, and operational infrastructure. A minority of Servitors may remain valuable as maintainers of the verification regime. The mass of implementers gets compressed further.

The Verdict

Deslop is useful engineering and a terminally inadequate economic defense. It can suppress obvious AI sludge, but that success accelerates the underlying process: more acceptable software produced with less human labor. It is not a shield against obsolescence. It is quality assurance for the machine that causes it.