Writing by Tatsuya Shimomotohttps://shimo4228.github.io/feed.xml2026-10-22T16:50:00+09:00Tatsuya Shimomotohttps://shimo4228.github.io/Using Claude Code to Rebuild Expertise I Don't Have, One Testable Rule at a Timehttps://dev.to/shimo4228/using-claude-code-to-rebuild-expertise-i-dont-have-one-testable-rule-at-a-time-eol2026-10-22T16:50:00+09:002026-10-22T16:50:00+09:00Dev.toWhen Consciousness Fully Stops, How Does the Mind Return?https://shimo4228.substack.com/p/when-consciousness-fully-stops-how2026-10-18T00:00:00+09:002026-10-18T00:00:00+09:00SubstackCan You Tell an Eval from a Review?https://dev.to/shimo4228/can-you-tell-an-eval-from-a-review-1b32026-10-13T09:00:00+09:002026-10-13T09:00:00+09:00Dev.toWhoa… My Claude Code Skill Descriptions Are Way Too Long…?!https://dev.to/shimo4228/whoa-my-claude-code-skill-descriptions-are-way-too-long-43l2026-10-09T20:39:00+09:002026-10-09T20:39:00+09:00Dev.toIs a README for Humans or for LLMs?https://dev.to/shimo4228/is-a-readme-for-humans-or-for-llms-22062026-10-09T09:00:00+09:002026-10-09T09:00:00+09:00Dev.toTurning Claude Code into a Writing Harness with Claude Mods and an Output Stylehttps://dev.to/shimo4228/turning-claude-code-into-a-writing-harness-with-claude-mods-and-an-output-style-1a732026-10-03T17:31:00+09:002026-10-03T17:31:00+09:00Dev.to"Is There Any Point to This?" Removing the Jev Plugins I Added to Claude Code After One Weekhttps://dev.to/shimo4228/is-there-any-point-to-this-removing-the-jev-plugins-i-added-to-claude-code-after-one-week-49eh2026-09-29T12:47:00+09:002026-09-29T12:47:00+09:00Dev.toWhy Does AI Writing Feel Like It's Floating?https://shimo4228.substack.com/p/why-does-ai-writing-feel-like-its2026-09-28T00:00:00+09:002026-09-28T00:00:00+09:00SubstackBefore Adding a Jev-Type Judgment Model, Read Your Generation Model's Probabilitieshttps://dev.to/shimo4228/before-adding-a-jev-type-judgment-model-read-your-generation-models-probabilities-5da02026-09-27T12:38:00+09:002026-09-27T12:38:00+09:00Dev.toMoving My Research Pipeline's Judgment Calls from an LLM to Jev, a Judgment-Only Modelhttps://dev.to/shimo4228/moving-my-research-pipelines-judgment-calls-from-an-llm-to-jev-a-judgment-only-model-4ncj2026-09-25T09:00:00+09:002026-09-25T09:00:00+09:00Dev.toWhat Does It Take to Reproduce Jev's Decisions Locally?https://dev.to/shimo4228/what-does-it-take-to-reproduce-jevs-decisions-locally-3i0n2026-09-24T09:00:00+09:002026-09-24T09:00:00+09:00Dev.toI Added Jev's Skill Router to Claude Code and Turned Back Just Before Rewriting the Skill Listinghttps://dev.to/shimo4228/i-added-jevs-skill-router-to-claude-code-and-turned-back-just-before-rewriting-the-skill-listing-34in2026-09-21T19:42:00+09:002026-09-21T19:42:00+09:00Dev.toHow Close to Opus Does Jev, a Model That Writes No Text, Get at Skill Selection in 0.3 Seconds?https://dev.to/shimo4228/how-close-to-opus-does-jev-a-model-that-writes-no-text-get-at-skill-selection-in-03-seconds-1nfj2026-09-21T10:02:00+09:002026-09-21T10:02:00+09:00Dev.toI Made the Top Model My Session Default, and One Heavy Implementation Plus Its Review Burned Through Fable's Usage Limithttps://dev.to/shimo4228/i-made-the-top-model-my-session-default-and-one-heavy-implementation-plus-its-review-burned-41822026-09-17T08:30:00+09:002026-09-17T08:30:00+09:00Dev.toPeople Inside the Business Don't Call It a 'Domain'https://dev.to/shimo4228/people-inside-the-business-dont-call-it-a-domain-3kb22026-09-16T09:00:00+09:002026-09-16T09:00:00+09:00Dev.toWhy an 11-Second Burst Never Showed Up in the Log That Gets Read Every Weekhttps://dev.to/shimo4228/why-an-11-second-burst-never-showed-up-in-the-log-that-gets-read-every-week-14nd2026-09-14T09:00:00+09:002026-09-14T09:00:00+09:00Dev.toI Deleted the LLM-Facing Architecture Docs I Had Committed 159 Times in 3 Months: Structure Goes to LSP, Reasons to ADRs, Diagrams to Humanshttps://dev.to/shimo4228/i-deleted-the-llm-facing-architecture-docs-i-had-committed-159-times-in-3-months-structure-goes-to-1a942026-09-05T19:58:00+09:002026-09-05T19:58:00+09:00Dev.toI Asked AI to Diagnose My Knowledge Blind Spots — 15 Days of Deadlock Moved in 85 Minuteshttps://dev.to/shimo4228/i-asked-ai-to-diagnose-my-knowledge-blind-spots-15-days-of-deadlock-moved-in-85-minutes-3j2n2026-09-02T19:19:00+09:002026-09-02T19:19:00+09:00Dev.toMy Dead-Code Scan Returned Zero, Then I Deleted 2,063 Lines: Detectors Measure References, Not Consumptionhttps://dev.to/shimo4228/my-dead-code-scan-returned-zero-then-i-deleted-2063-lines-detectors-measure-references-not-4o4e2026-08-31T08:30:00+09:002026-08-31T08:30:00+09:00Dev.toAfter Cutting My AI Reviews, I Put a Complexity Ceiling in Ruffhttps://dev.to/shimo4228/after-cutting-my-ai-reviews-i-put-a-complexity-ceiling-in-ruff-1hho2026-08-30T09:48:00+09:002026-08-30T09:48:00+09:00Dev.toI Cut My AI Review Chain From 6 Stages to 1: Breaking the Loop That Never Hits Zero Findingshttps://dev.to/shimo4228/i-cut-my-ai-review-chain-from-6-stages-to-1-breaking-the-loop-that-never-hits-zero-findings-1moi2026-08-29T07:20:00+09:002026-08-29T07:20:00+09:00Dev.toMy KPIs Improved. I Deleted My Homegrown LLM Judge Anywayhttps://dev.to/shimo4228/my-kpis-improved-i-deleted-my-homegrown-llm-judge-anyway-2n732026-08-27T19:49:00+09:002026-08-27T19:49:00+09:00Dev.toTwo Drafts Passed Every Eval and Both Were Hollow. Attach the Raw Transcript to the Ledger You Hand Your AIhttps://dev.to/shimo4228/two-drafts-passed-every-eval-and-both-were-hollow-attach-the-raw-transcript-to-the-ledger-you-hand-32p22026-08-21T22:20:00+09:002026-08-21T22:20:00+09:00Dev.toI Handed 41 Tasks to an AI Loop. The Bottleneck Was Judgment, Not Codehttps://dev.to/shimo4228/i-handed-41-tasks-to-an-ai-loop-the-bottleneck-was-judgment-not-code-23dp2026-08-20T19:25:00+09:002026-08-20T19:25:00+09:00Dev.toAI Review Kept Creating Work: Why I Deleted 4,541 Lineshttps://dev.to/shimo4228/ai-review-kept-creating-work-why-i-deleted-4541-lines-22ec2026-08-16T22:38:00+09:002026-08-16T22:38:00+09:00Dev.toCan Six-Month-Old AI Code Survive Today's Review? A 25-Bug Triagehttps://dev.to/shimo4228/can-six-month-old-ai-code-survive-todays-review-a-25-bug-triage-4o6d2026-08-14T06:02:00+09:002026-08-14T06:02:00+09:00Dev.toAfter I Ran Out of Things to Ask AI, One Desire Remainedhttps://shimo4228.substack.com/p/after-i-ran-out-of-things-to-ask2026-08-13T00:00:00+09:002026-08-13T00:00:00+09:00SubstackClaude Code's Memory Has No Vectors — Try ADRs Before Memory RAGhttps://dev.to/shimo4228/claude-codes-memory-has-no-vectors-try-adrs-before-memory-rag-4kik2026-08-06T09:00:00+09:002026-08-06T09:00:00+09:00Dev.toOn Embodiment in the Age of AIhttps://shimo4228.substack.com/p/on-embodiment-in-the-age-of-ai2026-08-04T00:00:00+09:002026-08-04T00:00:00+09:00SubstackGive Claude Code a Second Harnesshttps://dev.to/shimo4228/give-claude-code-a-second-harness-27of2026-08-03T09:00:00+09:002026-08-03T09:00:00+09:00Dev.toBanned from Wikidata Overnight — I Believed Every Edit Was Compliant, but All 109 Items Were Deleted as Promotionhttps://dev.to/shimo4228/banned-from-wikidata-overnight-i-believed-every-edit-was-compliant-but-all-109-items-were-57nl2026-07-30T08:32:00+09:002026-07-30T08:32:00+09:00Dev.toWhat Humans Should Approve Is Intent, Not the Diff — A Decision Table for Agent Approval Gateshttps://dev.to/shimo4228/what-humans-should-approve-is-intent-not-the-diff-a-decision-table-for-agent-approval-gates-1a3j2026-07-29T12:13:00+09:002026-07-29T12:13:00+09:00Dev.toclaude-security Beta: Cost, Output, Accuracy — 189 Agents, 2 Hours, Measuredhttps://dev.to/shimo4228/claude-security-beta-cost-output-accuracy-189-agents-2-hours-measured-1eji2026-07-28T09:00:00+09:002026-07-28T09:00:00+09:00Dev.toOpus 5 Changed How Rules Should Be Written — Audit Yourshttps://dev.to/shimo4228/opus-5-changed-how-rules-should-be-written-audit-yours-4fb42026-07-27T09:00:00+09:002026-07-27T09:00:00+09:00Dev.toherdr, a tmux for AI Agents — Until the Editor Disappearedhttps://dev.to/shimo4228/herdr-a-tmux-for-ai-agents-until-the-editor-disappeared-3hnn2026-07-21T09:00:00+09:002026-07-21T09:00:00+09:00Dev.toWhat Do My AI Agent's Logs Look Like in OpenTelemetry?https://dev.to/shimo4228/what-do-my-ai-agents-logs-look-like-in-opentelemetry-27ic2026-07-17T09:00:00+09:002026-07-17T09:00:00+09:00Dev.toFault Injection TDD Found 3 Silent Failures in My LLM Agenthttps://dev.to/shimo4228/fault-injection-tdd-found-3-silent-failures-in-my-llm-agent-3mld2026-07-16T09:00:00+09:002026-07-16T09:00:00+09:00Dev.toLLM-as-Judge Shouldn't Aggregate Scores: Binary Checks as Evidence, One Holistic Verdicthttps://dev.to/shimo4228/llm-as-judge-shouldnt-aggregate-scores-binary-checks-as-evidence-one-holistic-verdict-8222026-07-15T09:00:00+09:002026-07-15T09:00:00+09:00Dev.toBuilding an Autonomous Agent on an M1 Mac, by Choicehttps://dev.to/shimo4228/building-an-autonomous-agent-on-an-m1-mac-by-choice-5b5o2026-07-14T09:00:00+09:002026-07-14T09:00:00+09:00Dev.toWhy Did My Agent Decide That? 3 Observability Patternshttps://dev.to/shimo4228/why-did-my-agent-decide-that-3-observability-patterns-ami2026-07-13T09:00:00+09:002026-07-13T09:00:00+09:00Dev.toClaude Code from iPhone: Plugging 3 Holes in Remote Controlhttps://dev.to/shimo4228/claude-code-from-iphone-plugging-3-holes-in-remote-control-17cf2026-07-09T09:00:00+09:002026-07-09T09:00:00+09:00Dev.toLinters Can't Measure a Non-Code Asset's Value — I Built an LLM Stocktake for Ithttps://dev.to/shimo4228/linters-cant-measure-a-non-code-assets-value-i-built-an-llm-stocktake-for-it-4ng72026-07-08T09:00:00+09:002026-07-08T09:00:00+09:00Dev.toI Built a Skill for Easy Codex Reviews from Claude Codehttps://dev.to/shimo4228/i-built-a-skill-for-easy-codex-reviews-from-claude-code-4h892026-07-07T07:00:00+09:002026-07-07T07:00:00+09:00Dev.toYou Got It Right — But You Didn't Know Ithttps://shimo4228.substack.com/p/you-got-it-right-but-you-didnt-know2026-07-02T00:00:00+09:002026-07-02T00:00:00+09:00SubstackWhen Consciousness Fully Stops, How Does the Mind Return?https://shimo4228.substack.com/p/when-consciousness-fully-stops-how2026-06-14T00:00:00+09:002026-06-14T00:00:00+09:00SubstackWhere Does the Accountability Externalized by AI Go?https://shimo4228.substack.com/p/where-does-the-accountability-externalized2026-05-24T00:00:00+09:002026-05-24T00:00:00+09:00SubstackIs ReAct Needed in Production? — Separating Design and Operation Phaseshttps://dev.to/shimo4228/is-react-needed-in-production-separating-design-and-operation-phases-4ll02026-05-01T17:00:00+09:002026-05-01T17:00:00+09:00Dev.to(3) The LLM Workflow Quadrant Is Missing from Our Vocabularyhttps://dev.to/shimo4228/3-the-llm-workflow-quadrant-is-missing-from-our-vocabulary-n182026-04-30T17:00:00+09:002026-04-30T17:00:00+09:00Dev.toWhere ReAct Agents Are Actually Needed in Businesshttps://dev.to/shimo4228/where-react-agents-are-actually-needed-in-business-33do2026-04-29T13:12:00+09:002026-04-29T13:12:00+09:00Dev.toOrganic Growth and Content Integrity in an AI Writing Teamhttps://dev.to/shimo4228/organic-growth-and-content-integrity-in-an-ai-writing-team-1h672026-04-18T07:00:00+09:002026-04-18T07:00:00+09:00Dev.to