The Closed Track: Shrinking an AI Agent's Blast Radius Part 4
A code-running agent tool inherits its pod's blast radius. NetworkPolicy, a dropped ServiceAccount token and gVisor shrink it — and a blast-radius report shows which layers actually hold.
#agentic-ai
Driving Permits: Delegated, Short-Lived Authority for AI Agents Part 3
A signed task proves who wrote it, not who allowed it. RFC 8693 permits with sub, act, scope and a 120s TTL, checked by OPA at the MCP gateway.
License Plates: Giving AI Agents an Identity That Survives the Queue Part 2
mTLS ends at the queue. SPIRE-issued identities and JWS-signed tasks stop forgery and tampering. A k3d lab shows what they still can't stop.
The Safe Highway: Why Autonomous Agents Need a Road of Trust (and Why mTLS Isn't the Whole Road) Part 1
Autonomous agents on a bumpy network crash with no one to blame. mTLS paves the road, but it can't tell you who's driving. A k3d lab shows both.
The Frontier Has a Brake Pedal. The Floor Doesn't.
In June I asked whether Anthropic's warning was a brake pedal or a press release. The essay arrived — and it's a real mechanism, aimed one layer above where most of us actually ship agents.
Where Is AI Cost Really Going? A Practitioner's Math, Not a Pundit's Forecast
Over $1,200/month across Cursor and Claude, after already running Headroom and Ponytail. Here's why cheaper tokens aren't producing cheaper bills, and what that means for 2027 budgets.
Put Your Ego Aside and Start Using AI
The biggest thing stopping engineers from adopting AI isn't the tooling or the risk. It's the refusal to accept that an agent can out-produce them. Here's what that costs.
Beyond the Prompt: The Architecture of Loop Engineering
Prompt engineering micro-managed text. Context engineering hydrated it. Loop engineering builds the feedback loop the AI runs inside — and the token economy that makes or breaks it.
Anthroipcs announcement, a Brake Pedal, or a Press Release?
Anthropic wants the option to pause frontier AI development. A platform engineer reads the data behind the 'recursive self-improvement' warning — and the timing.
The AI Headcount Panic, Bad Bets, and Lost Knowledge
Companies across the globe—and Israeli high-tech especially—are mass-laying off in the name of AI. But is AI really the reason, or just the best available excuse?
The Infrastructure Wall: Why Your Agent Demo Died in Production
Everyone prototypes an AI agent in a weekend. Almost nobody ships it cleanly. Here's the wall you're about to hit — and how the platform is evolving to remove it.
ClawJacked, Axios, and the Autonomous Agent Problem
A practitioner's field notes on March 2026: OpenClaw's CVE flood, the Axios npm RAT, and why self-hosted autonomous agents are standing in the blast zone.
I Want a Personal Agent. I'm Not Running One Yet — Here's What Would Change That
Part 2: sandboxing with agent-sandbox, evaluating nanobot and nanoclaw, prompt injection realities, and the pre-flight checklist before I trust an autonomous agent.
Andrej Karpathy Just Made RAG Obsolete — And All You Need Is Three Folders
Andrej Karpathy dropped a paradigm-shifting gist on building personal knowledge bases with LLMs — no vector DB, no embeddings, just raw/wiki/output folders. Here's what it means for the rest of us.
The Agent Cost Wars — Updated: GLM-5, M2.7, and What the Leaderboard Actually Tells Us
A follow-up to my MiniMax M2.5 piece — challenging my own assumptions with fresh Artificial Analysis data, GLM-5, M2.7, and what this means for coders in 2026.
The $1,892 Agent: MiniMax M2.5 and the Dawn of Always-On Intelligence
MiniMax M2.5 achieves near-Opus 4.6 performance at 3% the cost. What this means for always-on agents, the SWE-bench, and the falling cost of intelligence.