D
Loading
LinkedIn articles
Syndicated pieces, hosted on LinkedIn.
Shorter writing that publishes first to LinkedIn. The Three-Body Problem series is the canonical home of the argument; these articles are entry points and follow-ups.
May 13, 2026
The Icing Problem Your AI Doesn't Know It Has
Most enterprise AI deployments skip safety evaluations entirely. Here's what capability, behavioral, and adversarial evals actually test, and why skipping them produces Therac-25-level failures.
Read on LinkedIn →
May 11, 2026
Ender's Rule: Dragon Army Never Fights the Same Way Twice
Pre-authorized compliance profiles that shift your security posture at adversary speed. Three profiles, clear transition authority, four readiness metrics.
Read on LinkedIn →
May 8, 2026
Moving Away from GitHub: The Trick Is Over
GitHub's April 2026 crisis, merge corruption, a critical RCE, and high-profile departures, makes the case for GitLab Ultimate as the enterprise DevSecOps platform.
Read on LinkedIn →
May 7, 2026
Bridge Crew, Not Autopilot: Why AI Copilots Work Best as Your Number One
GitLab Duo, not Microsoft Copilot, shows what an AI copilot looks like when it is platform-native. A Star Trek lens on what actually works.
Read on LinkedIn →
May 6, 2026
We Checksum Our Model Weights. We Don't Check What's Inside Them.
SHA-256 verifies the archive. SC-5 (Forensic Verification) confirms the weights inside. Here is why the difference matters for federal AI governance.
Read on LinkedIn →
May 4, 2026
Schoolhouse Rock Said the Bill Became a Law. Your Pipeline Still Isn't Ready.
The FY 2026 NDAA is converting AI guidance into enforceable mandates. Here is how the legislative supply chain works and why early movers win.
Read on LinkedIn →
Apr 30, 2026
The Minority Report AI Governance Is Missing
72% of enterprises think they govern their AI. Most don't. A policy that's never been attacked by an adversary isn't a policy; it's a hypothesis. Here's the pre-flight that changes that.
Read on LinkedIn →
Apr 27, 2026
Google Built the Orchestra. Nobody Tuned the Instruments.
Google, ServiceNow, and Microsoft announced multi-agent orchestration this week. Nobody is solving the context fidelity problem at handoff. Every re-transcription is lossy.
Read on LinkedIn →
Apr 23, 2026
The Model That Argues Back: Taming Claude Opus 4.7
Opus 4.7 treats specifications as negotiating positions. Four harness changes contain the variance without breaking what makes the model sharp.
Read on LinkedIn →