RC RANDOM CHAOS

Articles

Long-form writing on tech, culture, and the edges of the internet.

Our incident report was a vendor blog post
production-incidentsanthropic-api

Our incident report was a vendor blog post

Anthropic's invisible-guardrails apology misses the point: production agents need output contracts, audit ledgers, and sentinel checks, not model-default trust.

6 min read
Our repair agent patched the wrong file four times
claude-codepost-mortem

Our repair agent patched the wrong file four times

A Claude Code repair agent patched the wrong file four times in 31 hours. Why agents anchor on tracebacks, and the prompt rewrite that fixed it.

7 min read
Same commit, two builds, checksums that disagree
supply chain securitysoftware provenance

Same commit, two builds, checksums that disagree

A commit hash guarantees the repository, not the software. How build pipelines resolve trust once at the reference and inherit it against changed content.

7 min read
The scan isn't the story
ai agentsexecution context

The scan isn't the story

An AI agent with unchecked access bankrupted its operator. The failure was the execution context, not the scan. A breakdown of the boundary that never held.

8 min read
31 seconds per 100 lines
claude-codeproduction-incidents

31 seconds per 100 lines

38,412 lines of Claude-generated code, 11 incidents, one 11-day silent failure: why generation speed without verification slows your platform down.

7 min read
AWS Bedrock puts Anthropic inside your data path
aws bedrockanthropic

AWS Bedrock puts Anthropic inside your data path

AWS Bedrock's required data sharing with Anthropic redefines the trust boundary for third-party LLMs. What failed, why, and what must now be true.

8 min read
DeepSeek-R1 escapes its release boundary
DeepSeek-R1model release security

DeepSeek-R1 escapes its release boundary

DeepSeek-R1's open reproduction proves a release model with zero enforcement points: anyone can obtain and run the capability. A control failure briefing.

9 min read
Every commit swipes your badge at the door
identity-and-access-managementlateral-movement

Every commit swipes your badge at the door

Commits execute under identities. Unenforced IAM boundaries turn routine development into unowned access grants. What failed, why, and what must change.

7 min read
In January 2025, a hash passed for proof
systems driftmodel provenance

In January 2025, a hash passed for proof

The open reproduction of DeepSeek-R1 shows verification has no place inside the systems that consume model artifacts. Adoption ran on reference alone.

7 min read
Panic on a schedule
AI safetyGPT-2

Panic on a schedule

What the 2019 GPT-2 release panic predicted about GPT-4-era AI anxieties, and the misuse pattern that has repeated with every model since.

7 min read
PgDog's funding does not make it dangerous
ransomwareidentity-and-access

PgDog's funding does not make it dangerous

PgDog shifts ransomware to direct database infrastructure attacks. The enabling failure: identity and access controls that did not hold under exercise.

8 min read
The boundary did not hold
trust boundariesai agents

The boundary did not hold

An AI agent ran uncontrolled on a default Fedora setup. The failure was not the agent. It was trust assumed by default and enforced nowhere.

9 min read