DEV Community

Finley Zhou profile picture

Finley Zhou

Backend developer who loves Rust and distributed systems.

Location Tokyo, Japan Joined Joined on 
Do Not Let an Agent Patch Edit the Oracle It Must Pass

Do Not Let an Agent Patch Edit the Oracle It Must Pass

Comments
8 min read
Three Gates for Agent Patches: Frozen Flakes, Pinned Fixtures, Properties

Three Gates for Agent Patches: Frozen Flakes, Pinned Fixtures, Properties

Comments
7 min read
Merge Agent Patches Only When Two Test-World Receipts Match

Merge Agent Patches Only When Two Test-World Receipts Match

Comments
8 min read
Lock Behavior at HEAD Before an Agent Can Add Tests

Lock Behavior at HEAD Before an Agent Can Add Tests

Comments
7 min read
Put Property Checks on the Impact Cone, Not on the Edited Function

Put Property Checks on the Impact Cone, Not on the Edited Function

Comments
8 min read
CI Green After an Agent Patch Is Not an Inventory Diff

CI Green After an Agent Patch Is Not an Inventory Diff

Comments
8 min read
Tautologies Are Not Tests. Classify Agent Patches Before Merge

Tautologies Are Not Tests. Classify Agent Patches Before Merge

2
Comments 3
6 min read
Reject Agent Diffs That Widen skip, xfail, or Timeouts

Reject Agent Diffs That Widen skip, xfail, or Timeouts

Comments
9 min read
Quarantine Cannot Vote on an Agent Patch

Quarantine Cannot Vote on an Agent Patch

Comments
7 min read
Four Test Hunk Classes. Merge Only One With Production Code.

Four Test Hunk Classes. Merge Only One With Production Code.

Comments
7 min read
Diff the Oracle: Refuse Agent Patches That Weaken Assertions

Diff the Oracle: Refuse Agent Patches That Weaken Assertions

Comments
7 min read
Every New Branch in an Agent Diff Needs a Decision Row

Every New Branch in an Agent Diff Needs a Decision Row

Comments
8 min read
Assertion Loss Is a Regression. Score It in CI.

Assertion Loss Is a Regression. Score It in CI.

Comments
9 min read
Same-Session Tests Are Not Evidence. Isolate the Oracle.

Same-Session Tests Are Not Evidence. Isolate the Oracle.

Comments
7 min read
Score Assertion Weakening Before You Merge an Agent Patch

Score Assertion Weakening Before You Merge an Agent Patch

Comments
9 min read
Withhold the Oracle: Fixture Entropy, Metamorphic Checks, and a Flake Budget

Withhold the Oracle: Fixture Entropy, Metamorphic Checks, and a Flake Budget

Comments
8 min read
Oracle Order for Agent Diffs: Replay, Properties, Then a Flake Budget

Oracle Order for Agent Diffs: Replay, Properties, Then a Flake Budget

Comments
7 min read
Track Property Provenance Before You Merge an Agent Patch

Track Property Provenance Before You Merge an Agent Patch

Comments
8 min read
Agent Patches Evade Name Freezes. Bind Flakes to Failure Signatures.

Agent Patches Evade Name Freezes. Bind Flakes to Failure Signatures.

Comments
7 min read
Pin the Seed and Keep the Contract: A Merge Gate for Agent Patches

Pin the Seed and Keep the Contract: A Merge Gate for Agent Patches

Comments
8 min read
Pin Time, Maps, and IDs Before an Agent Patch Can Pass

Pin Time, Maps, and IDs Before an Agent Patch Can Pass

Comments
8 min read
Write-Block the Test Tree Before an Agent Patch Can Go Green

Write-Block the Test Tree Before an Agent Patch Can Go Green

Comments
7 min read
Silent Defaults Are the Real Agent Diff. Ledger Them Before Merge.

Silent Defaults Are the Real Agent Diff. Ledger Them Before Merge.

Comments
6 min read
Hide the Grading Suite: Agent Patches Need a Second Process

Hide the Grading Suite: Agent Patches Need a Second Process

Comments
8 min read
The Model Writes the Patch. Three Files Decide If It Merges.

The Model Writes the Patch. Three Files Decide If It Merges.

Comments
7 min read
Lock the Test Surface, Then Prove the Agent Patch With Relations

Lock the Test Surface, Then Prove the Agent Patch With Relations

Comments
8 min read
Seal the Oracle Before You Merge an Agent Patch

Seal the Oracle Before You Merge an Agent Patch

Comments
8 min read
Keep a Blind Fixture Set for Agent Patches

Keep a Blind Fixture Set for Agent Patches

Comments
7 min read
Unproven Is Not Green: Rank Agent-Patch Tests, Then Isolate Flakes

Unproven Is Not Green: Rank Agent-Patch Tests, Then Isolate Flakes

Comments
6 min read
Classify the Test Diff Before You Trust an Agent Patch

Classify the Test Diff Before You Trust an Agent Patch

Comments
7 min read
Split the Flake Before You Freeze It

Split the Flake Before You Freeze It

Comments
7 min read
Agent Patches Need a Sensor Queue, Not a Skip List

Agent Patches Need a Sensor Queue, Not a Skip List

Comments
9 min read
Measure Oracle Power Before You Merge an Agent Patch

Measure Oracle Power Before You Merge an Agent Patch

Comments
9 min read
Reject Agent Patches That Quiet the Test Suite

Reject Agent Patches That Quiet the Test Suite

Comments
8 min read
Split Mixed Agent Diffs: The Test Hunk Is Not a Witness

Split Mixed Agent Diffs: The Test Hunk Is Not a Witness

Comments
7 min read
Put a Write Boundary Between the Agent Patch and the Oracle

Put a Write Boundary Between the Agent Patch and the Oracle

Comments
8 min read
Run Only the Invariants the Diff Can Break

Run Only the Invariants the Diff Can Break

Comments
8 min read
Map Callers Before You Trust a Green Agent Patch

Map Callers Before You Trust a Green Agent Patch

Comments
9 min read
Metamorphic Checks for Agent Patches When You Have No Trusted Oracle

Metamorphic Checks for Agent Patches When You Have No Trusted Oracle

Comments
7 min read
A Receipt for Every Agent Patch: Seeds, Digests, and a Freeze File

A Receipt for Every Agent Patch: Seeds, Digests, and a Freeze File

Comments
7 min read
Green Tests Can Be a Weakened Oracle: A Three-Bucket Plan for Agent Patches

Green Tests Can Be a Weakened Oracle: A Three-Bucket Plan for Agent Patches

Comments
8 min read
Green CI Is Not a Boundary Test: Partition Tables for Agent Patches

Green CI Is Not a Boundary Test: Partition Tables for Agent Patches

Comments
9 min read
Name the Failure Class Before You Keep an Agent Patch

Name the Failure Class Before You Keep an Agent Patch

Comments
7 min read
Test-Inventory Diff as a Merge Gate for Agent Patches

Test-Inventory Diff as a Merge Gate for Agent Patches

Comments
7 min read
A Frozen Test Is Not Coverage: A Merge Gate for Agent Patches

A Frozen Test Is Not Coverage: A Merge Gate for Agent Patches

Comments
10 min read
Shadow-Compare the Agent Patch. Merge Only Classified Divergences.

Shadow-Compare the Agent Patch. Merge Only Classified Divergences.

Comments
8 min read
Don't Xfail the Flake. Journal Its Seed.

Don't Xfail the Flake. Journal Its Seed.

Comments
7 min read
Classify the Failure Before the Agent Edits a Test

Classify the Failure Before the Agent Edits a Test

1
Comments
7 min read
Diff the Test Suite, Not Just the Exit Code, After an Agent Patch

Diff the Test Suite, Not Just the Exit Code, After an Agent Patch

Comments
9 min read
The Compiler as Validator: When a Free Model's Wrong C++ Is Still Cheap

The Compiler as Validator: When a Free Model's Wrong C++ Is Still Cheap

Comments
4 min read
Case Study: A Free Model's 47th Test Input Crashed Our Parser

Case Study: A Free Model's 47th Test Input Crashed Our Parser

Comments
5 min read
Capture the Oracle Before the Agent Runs

Capture the Oracle Before the Agent Runs

Comments
8 min read
The 3 AM Restart: A 25-Line Watchdog a Free Model Wrote for My Free Server

The 3 AM Restart: A 25-Line Watchdog a Free Model Wrote for My Free Server

Comments
5 min read
Free AI Log Triage: Cut Weekly Review to 15 Minutes

Free AI Log Triage: Cut Weekly Review to 15 Minutes

Comments
5 min read
Score Test Hunks Before You Trust a Green Agent Patch

Score Test Hunks Before You Trust a Green Agent Patch

Comments
9 min read
Build a Write-Protected Test Plane for Agent Patches

Build a Write-Protected Test Plane for Agent Patches

Comments
7 min read
Assertion Deltas Are the Real Test Report for Agent Patches

Assertion Deltas Are the Real Test Report for Agent Patches

Comments
8 min read
Don't Skip Flakes After an Agent Patch. Demote Them.

Don't Skip Flakes After an Agent Patch. Demote Them.

Comments
8 min read
Spend a Freeze Slot Only After Property and Fixture Both Fail

Spend a Freeze Slot Only After Property and Fixture Both Fail

Comments
7 min read
A Shrinking Input Domain Is a Failed Agent Patch

A Shrinking Input Domain Is a Failed Agent Patch

Comments
8 min read
loading...