Blog archive

Agent count is a vanity metric. It tells you about volume, not value. Here is what I track instead after running a one-person AI fleet.

I run a one-person company on scheduled agents and gave almost none of them memory. They write to files instead. Here is why that wins.

Split a 70B model across multiple GPUs with llama.cpp. How --tensor-split, --main-gpu, and --split-mode work on a real consumer rig.

The cost gap between what an AI agent could cost and what it does cost is 40%. You close it at the call site, not in a dashboard. Here is how.

A 2026 Mem0 survey found 57-71% cross-user memory contamination across major agent frameworks. Here is why it happens and how to stop it.

JPMorgan turned on AI for 250k people. The quiet line is that the usage racks up fees. Here is how to control the bill before it arrives.

Anthropic filed for IPO at a $47B run-rate while 40% of enterprise customers report under 10% cost savings from Claude. Here is how to close that gap.

A repair agent in my own pipeline failed the same check 27 times in a row. Each try was a paid model call. Here is why uncapped retries quietly burn money, and the two-line fix.

Anthropic banned 832 accounts for AI-enabled attacks. What it means for teams running AI agents.

JPMorgan just switched on AI for 250,000 employees. The headline is workforce shift. The quiet story is enterprise AI cost, and why token spend runs away without controls.

Most AI advice tells you to ship more agents. Here is the honest opposite: the four times a plain script and a human beat an agent, learned running a fleet daily.

Copilot went usage-based and bills spiked. The fix is a runtime budget cap at the call site.

Uber caps every employee at $1,500/month per AI coding tool. The real fix is a per-identity cap in code, not a policy memo.

You set -ngl 99 and llama.cpp still pins the CPU — the flag isn't the bug. Here's the 30-second load-log check and the 5 real causes, ranked by how often they bite.

Prompt instructions are a request. API contracts are a wall. Why I moved my blog QA gate out of the prompt and into the server.

My blog repair loop chewed on a stale draft for 23 mornings and reported "blocked" every time. The fix was not a smarter retry. It was a TTL and a heal path.

Scheduled tasks exit 0 even when the work never happened. Here is the outcome layer I built on top of my agent fleet, and why it shipped before any new dashboard.

Google found the first AI-built zero-day in a planned mass-exploitation event. A builder's read on what changes for small operators running agents.

AI billing is shifting from seats and tokens to outcomes. If you cannot tie an agent run to a dollar of work, you are paying for vibes.

Every dashboard was green and zero blog posts went live. Exit codes tell you the job ran, not that the outcome happened. Here is how to check the real artifact instead.

A new open protocol lets AI agents register users with your app, no signup form. Here is how it works and what breaks.

Claude Opus 4.8 dropped May 28, 2026. Same price as 4.7, higher SWE-bench scores, and a model that flags its own mistakes. Here is what actually changed if you build AI agents.

MIT Tech Review says the AI-jobs hysteria is overstated. The real story is cost discipline, not displacement.

If Microsoft can't absorb agent inference costs, neither can you. Make the cap a config change, not a memo.