Engineering notes from the agent era

Most posts here are about what changes when the agent becomes a first-class collaborator: local code intelligence, testing systems that analyze behavior over time, knowledge tooling that compounds. The rest of the career (cloud, low-latency, mobile) shows up when something's worth writing down.

An overhead editorial photograph of a walnut desk. An open laptop shows a blurred code editor, beside fanned blank index cards, two pale-yellow sticky notes, a folded sheet with a pencil sketch of boxes connected by arrows, a graphite pencil, and a glass of iced cold brew coffee.

Oct 19, 2026·16 min read

How I Get Work Done With Agents (October 2026)

How I work with agents as of October 2026: architecture in docs first, sluice routing each change by shape (8 min fast, 90 min deep), magpie on every PR.

An editorial photograph split into two halves. On the left, a messy heap of hundreds of slightly different printed cards spills across dark slate. On the right, one short neat stack of identical cream cards sits under a brass paperweight on pale paper.

Oct 1723 min read

Do You Still Need to Read AI-Generated Code?

I asked three coding agents for the same five functions 900 times. A ticket gave a median 26 behaviours per function; a precise spec gave 2. Read the spec.

AI agents
code review
code comprehension
specifications
Claude Code
Codex

An editorial photograph of a vast, empty exam hall with rows of wooden desks receding into the distance. In the foreground one desk holds a finished stack of paper with a pencil laid on top, and its chair is empty.

Oct 519 min read

Your Interview Measures the Agent, Not the Candidate

I gave four interview formats to Claude and Codex 438 times, unaided. Every take-home correctness check was met. What separates candidates isn't in the code.

technical interviews
hiring
AI coding agents
take-home assignments
engineering management

An editorial photograph of a dark beetle pinned to a cream card under a glass dome on a dark wooden plinth in a natural-history museum. Rows of glass-fronted wooden cabinets and a hanging whale skeleton recede into soft focus behind it.

Sep 2822 min read

Your Agent's Tests Pass. The Bug Is in the Chat Log.

I planted bugs in 18 functions and asked Claude to write tests. Given only the code, the suites locked the bug in and still scored 0.95 on mutation testing.

AI agents
mutation testing
agent-written tests
Claude Code
AI engineering metrics

An editorial photograph looking straight down on a pale grey surface. A pencil worn to a short stub and an unused white eraser still in its paper sleeve sit beside a squared stack of blank cream sheets.

Sep 2115 min read

The Code Only Goes One Way: Why Agents Won't Delete

Across 2,115 agent edits to code files in my own sessions, 73.7% grew the file and 10.4% shrank it. A July 2026 paper names the mechanism: deletion avoidance.

AI agents
code quality
technical debt
Claude Code
AI engineering metrics

An editorial photograph of a worn brass spirit level in close-up, its bubble resting dead centre. The level lies across two very different stacks of books, thick cloth hardbacks on one side and frayed old paperbacks on the other, both softly blurred.

Sep 1923 min read

Is Your Code Context Good? The Benchmarks Can't Tell You.

Cursor benchmarked semantic code search at +12.5%, then deleted it. I ran 214 agent evals on my own repos: zero tasks discriminate on Opus 5.

code intelligence
AI agents
code search
evaluation
context engineering

An editorial photograph of a dark basement where a single beam picks out a brass pressure gauge on an old copper pipe. The gauge's needle points calmly upright while, further along the pipe, a drop of water hangs from a soldered joint.

Sep 1420 min read

Your Agent Recovers From Everything It Can See

Across 81,123 tool calls in 1,278 Claude Code sessions, 2.52% failed and 93.1% of those were plumbing. The agent recovered on the next call 91.2% of the time.

AI agents
agent observability
Claude Code
agent reliability
AI engineering metrics

An editorial photograph of the margin of an old book page in extreme close-up. Faint graphite pencil arrows, brackets, loops and underlines crowd the cream paper, while the printed column beside them dissolves into an unreadable blur.

Sep 715 min read

The Apprenticeship Was in the Comments

Anthropic put 52 junior engineers on an unfamiliar library. The AI-assisted half scored 50% against 67%, worst on debugging. Review taught that skill.

code review
junior developers
engineering management
mentorship
AI engineering metrics

An editorial photograph of an archive room crowded with towering stacks of aged paper and shelves of bundled files fading into soft focus. A brass-rimmed magnifying glass rests on the top sheet of the nearest stack only.

Aug 3118 min read

Merged Is Not Reviewed

Across 12,291 merged PRs in five OSS repos, code arriving for review rose 42% a month while review discussion per 100 lines fell 36%. Time-to-merge did not move.

code review
AI engineering metrics
agentic workflows
verification
delivery metrics

An editorial photograph of a long roll of blank cream paper unspooling in loose waves across a seamless cream backdrop into the distance. In the foreground sits a small, neatly squared stack of blank cards.

Aug 2422 min read

Agent Memory, Revisited: What Held Up and What Didn't

Zero reported benefit from transcript recall. Clean code cuts agent tokens 7-8% at unchanged pass rates. I audited 13 of my own agent-memory claims: 6 hold.

agent memory
Claude Code
agent skills
context engineering
AI agents

An editorial photograph taken straight down onto a pale birch board, with an open notebook filled with rows of pencil tally marks grouped in fives. A brass pocket stopwatch with a blank dial rests on the page beside a short pencil.

Aug 1730 min read

Measure the Machine, Not the Mood: How to Eval Your Own Week

A week of session eval found what a green build cannot: 58% of the clock was tool execution, re-reviews outnumbered reviews 10 to 7, and 17 of 23 flags wrong.

AI evals
session eval
agent observability
AI engineering metrics
agentic workflows

An editorial photograph of a dark wooden door standing ajar in a black hallway at night, a thin slit of warm lamplight spilling from the room inside. An old key is still in the lock, with a blank paper tag hanging from it on string.

Aug 1029 min read

Agents as the Attacker: The Threat Model Just Flipped

Hugging Face was breached by an autonomous agent framework across 17,000 logged actions. OpenAI later said it was their eval harness. The threat model flipped.

AI agents as attackers
agent security
credential scoping
coding agent sandbox escape
agentjacking

An editorial photograph in tight close-up of an iron key ring hanging from a nail on a plaster wall. Three old, rusted skeleton keys hang beside one new, polished brass key of nearly the same shape, with the wall falling into soft blur.

Aug 532 min read

LLM Model Deprecation: The Failures That Don't Error

Opus 4.1 retires today. Seventeen dated churn events in 61 days. The failures that error are the cheap ones. A pinning policy, a grep list, an eval gate.

LLM model deprecation
model version pinning
AI engineering
model migration
agent maintenance

An editorial photograph of a large open cardboard box on a seamless warm-grey backdrop, overflowing with crumpled brown packing paper and loose twine. Deep inside, almost lost in the packing, sits one tiny parcel tied with string.

Aug 232 min read

Your Harness Is the Cost Line, Not Your Model

Claude Code sends ~32,800 tokens before reading your prompt. Databricks changed harness, not model, and cut cost per task more than 2x at the same quality.

AI agent cost
Claude Code
token overhead
agent harness
AI engineering

An editorial photograph of an open, aged ledger with handwritten ruled entries on a dark desk, lit by a single brass lamp in an otherwise black room. A sealed envelope with a dark red wax seal lies across the middle of the pages, covering the key entry.

Jul 3118 min read

The Agent Audit Trail Paradox: Your Forensics Got Worse

Codex encrypted sub-agent task payloads and the readable agent audit trail went with it. Across 991 real delegations, 97.8% would fit an 8 KiB audit copy.

agent observability
AI agent security
audit trail
OpenTelemetry
AI engineering

An editorial photograph of a wall of wooden pigeonholes in an old sorting room, each holding its own envelope tied with string and carrying a blank paper tag. The sorting bench and floorboards behind fall into soft focus.

Jul 2828 min read

MCP Goes Stateless: The Migration Guide for Shipped Servers

MCP 2026-07-28 removes the handshake, the session id, and Sampling. Nothing breaks on July 28, which is the trap. Here is the migration, in blast-radius order.

MCP
model context protocol
protocol migration
platform engineering
AI engineering

An editorial photograph of a two-part sorting tray on a seamless cream backdrop, one half pale oak and the other walnut. The oak side is packed with blank cream cards, the walnut side holds only a few, and one card leans across the divider mid-sort.

Jul 2413 min read

Open Weight vs. Frontier: A Realistic 2026 Coding Verdict

Open-weight coders closed the gap to 5-10 points at a fraction of the price. But they ran 29% of tokens at under 4% of spend in 2026. Route, don't switch.

open weight
frontier models
LLM cost
self-hosting
AI engineering

An editorial photograph of a stone-floored pantry with long wooden shelves of identical jars of dark preserves, each sealed with a cream wax cap. In the sharp foreground, one jar's wax seal is cracked, while the rows behind fall into soft focus.

Jul 2118 min read

Your Agent's Tools Are a Supply Chain Now

A malicious MCP package ran clean for 15 releases, then a line BCC'd every email to an attacker. Agent tools are a supply chain. Secure them at install time.

MCP
supply chain security
AI agents
AI security
prompt injection

An editorial photograph of an old hand brace drill with a wooden handle and an empty brass chuck, lying on a seamless warm-grey backdrop. Two interchangeable auger bits rest beside it.

Jul 1920 min read

Keep the Harness, Change the Model: Run GPT Inside Claude Code

Keep Claude Code's tools, hooks, skills, MCP servers, and permissions while routing its agent harness to supported GPT models through a local proxy.

Claude Code
GPT
AI coding agents
developer tools
model routing

An editorial photograph of twelve identical brass pocket compasses with plain cream faces, laid flat in a neat grid on cream paper and seen from directly above. Every needle points the same way.

Jul 1816 min read

The Year the Coding UI Stopped Being a Text Box

In four months of 2026, six coding tools shipped the same parallel-agent command center. With Codex past 5M weekly users, the UI stopped being anyone's moat.

AI coding tools
Codex
Claude Code
agentic coding
developer tools

An editorial photograph split down the middle. On the left, a worn wooden carpenter's toolbox full of chisels, a hammer and a hand plane sits against a warm plaster wall. On the right, a pair of rusting tin-can telephones lies unused on a grey floor, joined by a slack string.

Jul 1517 min read

A2A vs. MCP: The Coordination Layer You Haven't Needed Yet

MCP crossed 97M monthly downloads; A2A hit 150+ orgs in a year. But most teams reaching for A2A just need in-process orchestration. Here's the decision tree.

A2A
MCP
AI agents
multi-agent systems
platform engineering

An editorial photograph of a raked gravel path curving between tall clipped yew hedges on a misty, overcast morning. At the end of the path stands a closed wooden garden gate between two stone pillars.

Jul 1020 min read

Block the Bad Agent in CI: Gate on Trajectories, Not Just Outputs

More than 1 in 5 GitHub code reviews now involve an agent. Pass@1 is the wrong merge gate. Gate on the trajectory, with a flake budget. Here is the CI policy.

agent evals
CI/CD
agent PR review
trajectory evaluation
AI agents

An editorial photograph of two matching dark oak boxes in a split frame. On the left, against dark charcoal plaster, one box is locked with a plain brass padlock through an iron hasp. On the right, against pale cream plaster, the other is held only by a loose paper ribbon bow, its lid sitting ajar.

Jul 720 min read

What Actually Stops Prompt Injection: Defenses vs. Theater

A 2026 lab study bypassed 12 prompt-injection defenses at over 90% attack success. Detection is theater; the durable wins are architectural. Here is what holds.

prompt injection
AI agent security
prompt injection defenses
CaMeL
AI agents

An editorial photograph of a perfectly still woodland pond under a grey overcast sky. The glassy water mirrors the clouds, the bare trees and the dry reeds that line its banks.

Jul 415 min read

Narcissus at the Keyboard

Sycophancy is structural, not a bug: models are trained to reflect you back. Why a mirror can't be a collaborator, and how to build the friction back in.

AI sycophancy
RLHF
AI agents
coding agents
LLM evaluation

An editorial photograph of the jaws of an aged brass vernier caliper closed on a small pale wooden block, shot close up. The caliper's beam recedes into soft blur against a plain grey background.

Jul 220 min read

Benchmarks Stopped Measuring Intelligence. Evaluate Models Yourself.

The same model scores 80.9% on SWE-bench Verified and 45.9% on contamination-resistant Pro. A leaderboard number isn't a capability. Build your own eval.

LLM evaluation
AI benchmarks
SWE-bench
LLM-as-judge
model selection

An editorial photograph of three pale fibre ropes reaching in from different directions against a black background and twisting together into a single rope. The joined rope ends in a short black fuse, caught by one low side light.

Jun 3020 min read

The Lethal Trifecta: Why Your Coding Agent Is a Loaded Gun

Four AI products leaked private data to prompt injection in eight days, and attacks top 85% against the best defenses. The lethal trifecta is buildable.

prompt injection
MCP security
lethal trifecta
AI agents
indirect prompt injection

An editorial photograph of an antique brass balance scale alone on a seamless cream backdrop. Each pan holds an identical blank card, yet the beam rests tilted with one pan hanging lower than the other.

Jun 2718 min read

LLM-as-Judge Is the New Flaky Test

A 2026 RAND harness found no LLM judge is uniformly reliable; trivial rewrites flip verdicts. Here is the discipline that makes a green eval trustworthy.

LLM-as-judge
AI evaluation
benchmarks
SWE-bench
AI agents

An editorial photograph of a misty railway yard on an overcast morning. An old iron hand lever stands beside a track switch in the foreground, where most of the rails curve off toward a short siding ending at a timber buffer stop while one line runs on to the horizon.

Jun 2517 min read

The Router Pattern: When to Downshift the Model

A 5-25x cost spread separates Haiku from Opus, yet most agent turns pay top-tier rates. A per-call routing decision tree, and where downshifting breaks.

AI agents
model routing
LLM cost
Claude Code
agent pipelines

An editorial photograph of three lidless light oak boxes nested one inside another, seen from directly above on dark grey slate. A small brass key lies in the innermost box.

Jun 2318 min read

Sandboxing Coding Agents: The 9-Second Argument for Isolation

A Cursor agent deleted PocketOS's prod database and backups in 9 seconds. Prompt rules aren't a boundary; the kernel is. An isolation ladder for coding agents.

AI agents
sandboxing
security
Claude Code
DevOps

An editorial photograph of a thick pile of rough-edged cream pages tied with twine on a seamless warm-grey studio backdrop. Beside it a single crisp blank white card stands upright in a small brass bulldog clip.

Jun 2017 min read

Does Your CLAUDE.md Actually Help? The Research Says Maybe Not

An ETH Zurich study found context files cut coding-agent success up to 2% and raised cost 20-23%. Field data disagrees. When your CLAUDE.md earns its tokens.

Claude Code
AGENTS.md
context engineering
AI agents
developer productivity

An editorial photograph of two notebooks seen from directly above on a pale grey surface. On the left lies a worn brown leather notebook bulging with frayed, loose pages, and on the right a new notebook lies open to blank cream pages with a single folded blank card on its right-hand page.

Jun 1819 min read

The Session Handoff: When Your Attention Budget Is Spent

Anthropic frames context as an attention budget. Lost-in-the-middle costs 30%+ retrieval accuracy. Three axes of handoff, five triggers, one cut.

AI agents
context engineering
Claude Code
session handoff
attention

An editorial photograph of a dark room at night where a single black desk lamp lights an open ring-binder runbook on a dark desk. A small brass service bell sits at the edge of the light, and the rest of the room falls into shadow.

Jun 1320 min read

Agent SRE: Oncall and Escalation for Coding Agents

Anthropic's Apr 23 2026 postmortem took 6 weeks to detect. Five SRE primitives map to Claude Code hooks and turn agent oncall into a shipped artifact.

AI agents
SRE
oncall
Claude Code
hooks

An editorial photograph of a tall wooden rolling ladder hooked onto a brass rail against floor-to-ceiling dark wooden bookshelves in an old library. The aisle runs back past stone columns and a gallery level into soft, hazy daylight.

Jun 1124 min read

The Promotion Ladder: Prompt, Skill, Hook, Tool

Four rungs trade flexibility for determinism. AGENTIF puts the best model at under 30% adherence; well-designed schema cuts MCP cost 99.9%. Match the rung.

skills
hooks
mcp
agents
claude-code

An editorial photograph of two hands passing a small folded cream note against a plain blurred grey backdrop. The hand on the left has already opened and is letting go, touching the note with one fingertip, while the hand on the right holds only its far corner.

Jun 724 min read

The Handoff Problem: When to Take the Keyboard Back

Developers use AI in 60% of work but fully delegate only 0 to 20% of tasks. Four named cues for the moment you reclaim: drift, scope, novel error, 80%.

autonomy
agents
claude-code
AI pair programming
agentic coding

An editorial photograph of a narrow stony footpath crossing misty upland moorland, marked by a line of stacked stone cairns. The nearest cairn stands sharp in the foreground and each one beyond grows smaller until the last fades into low cloud.

Jun 522 min read

Long-Running Autonomous Agents: Drift, Checkpointing, Recovery

METR pegs Opus 4.6 at a 14.5-hour 50%-time-horizon. Pass@1 collapses 24 points on long tasks. Drift as eval, checkpoint on threshold, fork not restart.

AI agents
Claude Code
agent reliability
long-horizon agents
agentic coding

An editorial photograph of a tall, squared stack of plain cream file folders on a seamless cream studio backdrop, set to the right of a wide empty frame. Two loose folders rest slightly askew on top of the stack.

Jun 316 min read

Cache-Aware Prompting: Engineering for 90%+ Hit Rate

ProjectDiscovery moved from 7% to 84% cache hit rate without changing the model. The discipline, the workload taxonomy, the five named failure modes.

AI engineering
prompt engineering
developer productivity
Anthropic
cost optimization

An editorial photograph of a vast, empty old boiler hall under overcast skylights, where a long tiled wall carries dozens of identical brass pressure gauges receding into the distance. One brass gauge on a pipe in the foreground is in sharp focus.

Jun 119 min read

Agent Cost Observability: From Personal Token Budget to Team-Wide

98% of FinOps teams now manage AI spend, up from 31% in 2024. Solo ccusage solved one developer; here is the three-axis, two-cap recipe for 200.

FinOps
AI engineering
developer productivity
observability
platform engineering

An editorial photograph of hand tools hanging on a dark wooden pegboard in a woodworking workshop, including a block plane, chisels, an awl, a mallet, a spokeshave and a try square, each with a blank cream luggage tag tied on with string. A worn workbench with a vice and curls of shavings falls away into soft focus behind.

May 3118 min read

Tool Design for Agents: Schema Is the Prompt

97% of MCP tool descriptions have at least one code smell. 56% fail to state purpose. The description is the prompt your agent reads to pick a tool. Here is the rubric.

AI agents
MCP
developer tools
prompt engineering
platform engineering

An editorial photograph split down the middle into two halves. On the left, a single closed oak card-catalogue drawer with a brass pull and an empty label frame sits in flat grey light. On the right, blank cream index cards pinned to cork with brass pins are joined to one another by taut white string.

May 3022 min read

The Project Graph: What Agents Need That Filesystems Can't Give

40 questions, two large repos, three LLM judges. Code-intelligence: judge 7.12 vs default 6.30 (+0.82), 29% faster, +8% tokens. Cites 50% vs CodeGraph's 32%.

AI agents
developer tools
MCP
code intelligence
platform engineering

An editorial photograph, shot from directly above, of five cream ceramic cups in a row on dark grey slate, each failing differently. One has a chipped rim, one a hairline crack, one brims over into a puddle, one sits empty, and the last lies tipped on its side beside a small spill.

May 2821 min read

The Five Failure Modes of Autonomous Coding Agents

Five named failure modes for autonomous coding agents, each with a real 2026 incident, a detection signal, and a retro template you can drop into CLAUDE.md today.

AI agents
incident response
Claude Code
evals
hooks

An editorial split-frame illustration: a bright, friendly desk with a laptop running a finished app on the left labelled localhost, separated by a thin vertical line from a darker server room with a billing alarm light glowing red on the right labelled production, used as a metaphor for the gap between an AI-built MVP and a deployed production application.

May 2722 min read

From Localhost to Production: The Handoff Brief for AI-Built Apps

45% of AI-generated code ships OWASP vulns. 380K vibe-coded apps public right now. The seven-gap handoff brief for builders and engineers.

AI engineering
vibe coding
production
security
developer experience

An editorial still-life of a three-tier workshop tableau on aged paper: a wooden letterpress typecase drawer of hundreds of identical metal sorts at the base, a small row of pre-cast composite component blocks in the middle, and a single completed galley page on top marked with handwritten proofreader's annotations, used as a metaphor for the layered substrate beneath AI-generated UI: Tailwind atoms, shadcn primitives, and AI-edited compositions

May 2517 min read

UI Libraries vs AI-Generated Components: The Tailwind Substrate

Tailwind 51%, v0 4M users, shadcn passed Chakra. The library-vs-AI debate is the wrong frame: substrate placement is the right one. The four-quadrant framework.

frontend
AI engineering
Tailwind
shadcn
developer productivity

An editorial illustration contrasting a wall of identical generic boxes labelled API mirror with a smaller set of differently-shaped purpose-built tools labelled tool shape, used as a metaphor for two opposing approaches to MCP server design

May 2320 min read

MCP Server for Your Codebase: Tool-Shape, Not API-Mirror

Cloudflare's first MCP server would have eaten 1.17M input tokens. Their redesign got it to roughly 1,000. Here is the framework, applied to a codebase server.

MCP
model context protocol
AI engineering
developer tools
platform engineering

An editorial photograph of a brass plumb bob hanging dead still on a thin cord in a dark, empty space, lit by a single narrow beam from above. Its point hovers just over a faint chalk cross on a cracked dark stone floor.

May 2115 min read

Claude Code Hooks: The Only Deterministic Substrate

The best frontier model follows under 30% of agentic instructions perfectly. Hooks run as code on every matched event regardless. Here is the substrate map.

Claude Code
AI agents
hooks
policy enforcement
agentic coding

A test bench with four labelled assertion gauges (skill trigger, subagent spawn, hook firing, MCP reachability) wired to a Claude Code session, used as a metaphor for control-plane evals against agentic machinery

May 1918 min read

Agent Evals: A Test Suite for Your Claude Code Setup

Observability says what happened. Evals say if the right thing happened. 89% ship the first, 52% the second. Four control-plane evals for Claude Code.

Claude Code
AI agents
evals
agentic coding
developer productivity

An editorial photograph of five aged brass keys hanging on small brass hooks across a textured off-white plaster wall, graded in size from smallest to largest. A worn wooden rubber stamp lies on its side on a narrow ledge below, softly out of focus.

May 1915 min read

The Permission Prompt Is Dying in AI Coding Agents

Claude Code users approve 93% of prompts. For AI coding agents, prompt walls failed as governance; safety is policy: allow, gate, block, log.

Claude Code
AI agents
permissions
policy enforcement
agentic coding

An editorial photograph of a small squat brass inkwell on a seamless cream backdrop, with five wooden dip pens crowded into its single narrow mouth, nibs down in the ink. A thin trickle of black ink runs down the rim, and the rest of the frame is empty.

May 1822 min read

Stdio MCP Doesn't Scale: Dropping 3,662 Lines for a Daemon

Five subagents across three repos loaded 2.6 GB of duplicated embedding models. v4 deleted the stdio path; the daemon shares everything. Here is the migration.

MCP
model context protocol
developer tools
AI engineering
platform engineering

Top-down photo of an open notebook spread. The left page is titled Mythos Audit May 11 2026 and shows five handwritten entries with four of them crossed out and one tagged low. The right page is titled AI-Coded CVE Log Nov 2025 to May 2026 and lists CVE identifiers with their CVSS scores. A red mechanical pencil rests across both pages along a thin red horizontal rule, used as a metaphor for the dual-angle read on AI as both bug-finder and bug-creator.

May 1618 min read

Claude Mythos vs. the CVE Surge: AI Security in May 2026

On May 11 curl's Daniel Stenberg called Anthropic's Mythos report mostly marketing. The same six months delivered the CurXecute RCE, the Claude Code chain, and a 35-CVE March.

AI security
CVE
Claude
Copilot
AppSec
developer productivity

A workshop wall with four labelled drawers (working, episodic, semantic, procedural) feeding a Claude Code session, used as a metaphor for the four-memory architecture mapped onto MEMORY.md, JSONL transcripts, the wiki, and skills

May 1517 min read

Agent Memory Architecture: Four Memories, Four Fixes

200K-context models rot by 50K tokens. Coding agents hit 150K in 35 minutes. Map four memories onto Claude Code: MEMORY.md, .remember, JSONL, skills.

Claude Code
AI agents
memory architecture
context engineering
agentic coding

An editorial illustration of two pools connected by a pipe: the larger left-hand pool is labelled Subscription Claude Code (interactive); the smaller right-hand pool is labelled Programmatic Credit at API rates; a vertical red bar labelled June 15 2026 severs the pipe between them, used as a metaphor for the split between interactive and programmatic Claude usage.

May 1516 min read

Anthropic Just Metered the Agent SDK: What Breaks on June 15

On May 13 Anthropic split Claude subscriptions into interactive and programmatic pools. Power users call it a 25x cost cut. Here is the strategic read.

AI engineering
Claude
Codex
agent SDK
developer productivity

An editorial illustration of a four-dial dashboard with three of the dials visibly disconnected from the underlying meters they claim to read, used as a metaphor for DORA's four classic metrics drifting from the system they were built to measure

May 1319 min read

DORA in the Agent Era: Three Metrics Stop Measuring

DORA's four metrics measured human-paced delivery. With agents writing 46% of code and review time up 441% YoY, three no longer measure what they claim.

DORA metrics
AI engineering
developer productivity
DevEx
engineering management

A single index card pinned to a workshop wall with seventeen lines of failing test handwritten on it, used as a metaphor for the failing test as the durable artefact for follow-up agent work

May 1115 min read

Agentic TDD: When the Failing Test Is the Spec

Spec-driven was last week's new feature. Today's spec: 17 lines of failing test. Artefact-driven TDD for follow-up agent work, against the 1.7x AI-issue rate.

Claude Code
AI agents
TDD
agentic coding
developer productivity

A workshop bench with three labelled drawers (brainstorm, design, plan) feeding a single output, used as a metaphor for the durable artefact pipeline that produces an agent-executable plan

May 1018 min read

Spec-Driven Agent Development: Brainstorm, Design, Plan

PR 23 told its reviewers to use MCP. They didn't. Per-agent tool calls jumped from 0.6 to 2.6 after a design doc surfaced the wiring bug prompts hid for weeks.

Claude Code
AI agents
spec-driven development
agentic coding
developer productivity

A wooden card-index drawer on a warm linen surface, packed with cream archival cards and a single card tilted forward, a physical metaphor for progressive disclosure

May 819 min read

Agent Skills: Progressive Disclosure That Actually Scales

Naive skill loading costs roughly 22x more tokens than progressive disclosure, and the attention math gets worse with every model upgrade. The pattern, the catalog math, and the authoring mistakes that break it.

agent skills
Claude Code
context engineering
progressive disclosure
AI agents

An overhead view of a precision manufacturing floor with three parallel conveyor lanes, used as a metaphor for the wave-based dispatch pattern in subagent-driven development

May 815 min read

Subagent-Driven Development: How to Fan Out a Feature Build

Subagents fan out feature builds at ~15x token cost. Wave dispatch, a frozen plan, and the five failure modes specific to subagent-driven development.

Claude Code
AI agents
subagents
developer productivity
agentic coding

A stack of well-worn engineering books on a desk in warm daylight, suggesting the durable layer beneath any paradigm

May 615 min read

Engineering That Outlasts the Paradigm

Trust in AI accuracy hit 29% the same year vibe coding became Word of the Year. Both numbers describe the same mistake. Engineering outlasts the paradigm.

AI agents
software engineering
agentic coding
career
thought leadership

A flock of starlings rendering an emergent coordination pattern across an open landscape, used as a metaphor for parallel subagent dispatch

May 416 min read

Subagent Patterns: When to Spawn vs Stay In-Context

Multi-agent burns 15x more tokens than chat. Five-question decision tree, 2026 token math, and three reproducible failure modes for Claude Code subagents.

Claude Code
AI agents
subagents
developer productivity
agentic coding

Paper with the line 'Use REST.' intact above red redaction bars covering the rest of the paragraph.

May 114 min read

We Tried to Cut Claude's Output 50%. We Got 5%. So Did Anthropic.

We aimed for 50% Claude output compression. We hit 4.7%. Anthropic hit the same wall and reverted at 3%. Here is the data and the failure mode.

claude-code
llm-output-compression
prompt-engineering
claude-skills
anthropic

An editorial illustration of a code repository as a layered query graph with typed signatures, file boundaries, and import edges traced by a precise probe

Apr 2914 min read

Your Codebase Is the Agent's Operating Environment

Frontier agents hit 90% on SWE-Bench Verified and 21% on SWE-EVO. The variable is the shape of the codebase, not the size of the model.

AI agents
monorepo
code intelligence
code graph
developer tooling

A clean editorial illustration of a code diff flowing through a fine-mesh sieve catching mechanical defects, then through a brass jeweler's loupe inspecting one architectural piece

Apr 2617 min read

AI Reviews the Diff. Humans Review the Decision.

AI code-review adoption tripled to 51.4% in 2025, but 31% more PRs now merge unreviewed. Honest market scan, security posture, and a Claude Code DIY recipe.

AI PR review
AI code review
Claude Code
GitHub Actions
developer tooling

A clean editorial illustration evoking record, replay, diff, and judge as a temporal evaluation loop

Apr 2517 min read

Backtesting AI Agents: Replay to Catch Regressions

54% of enterprises ship AI agents in production. Most cannot tell when a CLAUDE.md edit silently regresses behavior. Backtesting is the missing discipline.

agent evaluation
backtesting
Claude Code
AI agents
regression testing
LLM as judge

A 2x2 decision matrix mapping the four context surfaces (CLAUDE.md, skills, memory, MCP) against frequency and stability axes

Apr 2515 min read

Context Engineering in Practice: Where Does Each Piece Go?

Context engineering became the #1 2026 skill shift. Anthropic's research notes context exhibits n² token relationships. Here's the per-surface decision framework.

context engineering
Claude Code
MCP
AI agents
developer productivity

A sleek, conceptual representation of a stacked architectural infrastructure system

Apr 1827 min read

Treat AI as a Team Member, Not a Chat Window

84% of developers use AI, 46% distrust it. The right scaffolding (constitution, skills, memory, MCP, subagents) turns an assistant into a team member.

AI agents
developer productivity
Claude Code
MCP
team workflows

A premium, sophisticated abstract representation of a glowing technical dashboard or progress ring tracking data tokens

Apr 1611 min read

How to Track Claude Code 5-Hour Window Usage

40.8% of devs use Claude Code, but the 5-hour window is opaque. Build a local dashboard that parses transcripts, estimates your token budget, and rolls up team-wide cost via Grafana Loki.

claude-code
token-usage
developer-tools
ai-coding
cost-tracking

Abstract 3D visualization of code structure as a glowing interconnected graph of symbols and relationships

Apr 925 min read

Your AI Agent Is Flying Blind Without Local Code Intelligence

84% of developers use AI tools but 46% distrust the output. Three on-device models, 32 MCP tools, 9.93/10 relevance, and zero source code leaving your machine.

local code intelligence
AI agents
MCP
code search
developer tools

A dynamic, high-tech abstract visualization of a glowing interconnected wiki graph forming from raw data blocks

Apr 812 min read

Building an LLM Wiki: From Karpathy's Gist to a Working CLI

I turned Andrej Karpathy's LLM wiki concept into a Bun CLI (~500 lines of TypeScript) that automatically builds a persistent knowledge base from Claude Code sessions, files, and URLs.

llm
cli
knowledge-management
claude-code
bun

A clean, minimal editorial illustration of a data timeline and temporal behavior patterns

Apr 215 min read

How Do You Test Systems That Analyze Behavior Over Time?

Backtesting borrows from quant finance to catch temporal bugs unit tests miss. Poor US software quality costs $2.41T per year. Here's the technique.

backtesting
software-engineering
data-pipelines
temporal-data
regression-testing
synthetic-data
developer-tooling