sessions in my agent runtime are first-class objects. auto-named, resumable with full context via rolling summary, cross-device (desktop to CLI).

multi-agent collaboration patterns i'm building as first-class primitives: fan-out, pipeline, debate, orchestrator-worker.

no compaction in my agent runtime. instead of summarizing old messages and throwing them away, i use a sliding window plus a rolling summary. recent messages in full fidelity, older ones replaced by a continuously updated structured summary.

the runtime i'm building is model-agnostic. every agent definition declares a preferred model and a fallback chain. if claude hits a rate limit, it falls back to gpt 5.4 or glm 5.1.

experimenting with building my own agent runtime. i've customized the heck out of claude code, but it's locked to one model family and missing features i want. figured the best way to learn what i need is to build it myself.

exploring behavioral profiles in my agent runtime. instead of "code mode" vs "chat mode," each profile defines tool access, approval gates, and memory behavior. the UI adapts; what's exposed changes based on the profile.

librarium is a cli tool i built and maintain that fans out research queries across different providers. I just released v0.1.3 with a handful of new providers including firecrawl github.com/jkudish/lib...

claude code's built-in context tools are good. really good. but they're per-session, per-project. they don't compound across your whole practice. tomorrow's newsletter is about the platform i'm building on top to fix that. jkudish.com/newsletter

most developers write a CLAUDE.md, import a few skills, and stop there. i didn't stop there. wrote about what i built on top in tomorrow's newsletter. subscribe at jkudish.com/newsletter

i stopped thinking of AI as a coding assistant. it's a full implementation partner now. i describe the problem. it asks questions. i approve the plan. it writes the code, tests, opens the PR. my job is the decisions.

treat your CLAUDE.md as a table of contents, not an encyclopedia. short entry point with pointers to deeper docs. architecture in one file. conventions in another. agents navigate better with a map than a manual.

some teams are still debating whether to adopt AI. early ones are experimenting with ways to get reorganized around it. what i'm seeing as successful patterns: simple tasks go straight to agents. manageable tasks get a few feedback cycles. complex tasks stay with engineers.

what does your AI coding setup look like right now? single agent sessions? parallel worktrees? custom commands? something else entirely? curious how people are structuring their workflows beyond "open chat, type prompt."

the best way to make AI agents reliable isn't better prompts. it's fewer choices. test suites, linters, type systems, architectural constraints. every guardrail you add removes a degree of freedom where the agent could fail. agents don't need more intelligence. they need less room to be wrong.

every friday: run a cleanup agent across my repos. scans for stale docs, dead code, outdated patterns from the week. opens fix-up PRs automatically. agents replicate what exists, including the bad parts. regular garbage collection keeps drift from compounding.

I'm hiring a Video Editor & Producer to work with me at Builder Methods. Know someone? You? Looking for a creative partner to take this to the next level — storytelling, custom visuals, motion graphics. I value taste and creator chops above tactics. buildermethods.com/jobs/video-e...

implementation cost is going to zero. so what's left? vision. taste. judgment. the cognitive demands didn't shrink. they concentrated.

about 30% of my dev time goes to meta-work now. custom commands, agents, feedback loops, memory systems. not shipping features. building the system that ships features. every piece compounds into the next. and eventually, i want other builders to be able to use it too. more on that soon.

the pragmatic engineer surveyed 900+ engineers. 56% now do 70%+ of their work with AI. claude code went from zero to the #1 AI coding tool in 8 months. this isn't early adoption anymore. also who are the other 44% and what's the 30%? am i weird for being at like 99%?

SaaS isn't dying. but it's no longer the default answer to every business problem. Looks like a growing number of businesses are hiring builders to create internal tools instead of buying another subscription. If I can be of help, reach out!

last week i shipped across 6 projects in ~35 hours. not because i worked harder. because AI handled most of the execution. the leverage is real. but i'm still the bottleneck for every prompt. writing about that tension in this week's newsletter. subscribe: jkudish.com/newsletter

built a /deep-review command: fans out code review to 4 AI models in parallel. if 2+ flag the same issue, severity gets bumped. auto-fixes critical and high findings before the PR exists. built on @aarondfrancis's counselors package.

what's the one thing in your AI coding setup that compounds the most? for me it's context. CLAUDE.md files, auto-memory, project-specific configs. every session leaves behind knowledge that the next session builds on. the AI gets better at my codebase without me doing anything extra.

the more i automate my dev workflow, the more my job looks like product management. define the problem. clarify requirements. approve the plan. review the output. the engineering skills didn't become less valuable. they just moved from typing code to systems design.

claude code PostToolUse hooks are underrated IMO. i added a 17-line bash script that detects gh pr create and automatically starts the code review loop. no manual step between "PR created" and "reviewer comments handled." code.claude.com/docs/en/hooks

do you want AI to cite your brand? own the definition. find the specific terms in your niche that don't have clean, authoritative explanations anywhere. write the best one. structure it so AI can extract it cleanly. I call this the "dictionary definition heist."

three layers to my agentic memory system: 1. raw transcripts. every conversation stored as source of truth. 2. compressed observations. what the agent noticed and extracted. 3. normalized truths. facts with confidence scores and validity windows.

most AI conversations start from zero. you open a new chat and the AI has no idea what you worked on yesterday. I built a memory system so mine doesn't work that way anymore. three layers: raw transcripts, compressed observations, normalized truths with confidence scores.

i've been experimenting with different AI memory systems. i wanted something that can store observations, connect entities, and retrieve context across sessions. my first instinct was to reach for a hosted vector database. that's what everyone recommends. 🧵

caught up with the latest development happenings in WordPress over the weekend. some of my clients are heavily invested in the ecosystem and it's important to stay current. the biggest shift: the WordPress Interactivity API (IAPI). 🧵

icymi: i published my first npm package this weekend, 🏛️ Librarium it's an open source CLI tool that lets you fan out research queries to multiple search and deep-research APIs in parallel for human or agent consumption. let me know what you think! github.com/jkudish/lib...

🪶 announcing plume — a Laravel package for the X API v2. if you're building anything with the X API in Laravel, i hope this becomes the canonical implementation. give it a try and feel free to star/share 🙏 github.com/jkudish/plume

First time sharing my work on Hacker News. I posted about my new CLI tool Librarium. news.ycombinator.com/item?id=471... github.com/jkudish/lib... Appreciate an upvote if you can!

🏛️ Introducing Librarium Fan out research queries to multiple search and deep-research APIs in parallel. Results are collected, normalized, and deduplicated into structured output that your agents can consume or write to markdown. github.com/jkudish/librarium

🏛️ Introducing Librarium Fan out research queries to multiple search and deep-research APIs in parallel. Results are collected, normalized, and deduplicated into structured output that your agents can consume or write to markdown. github.com/jkudish/librarium

the real unlock with AI agents isn't just speed. it's parallelism. i work faster AND on more things at once.

what's one thing you automated this week that you used to do manually? for me: new errors on most of my projects automatically create an AI driven investigation and PR for me to review and approve.

i've been deep in SEO/AEO/GEO across several sites lately. if those terms are new: - SEO: rank on google - AEO: get cited by ChatGPT, Perplexity, Gemini - GEO: show up in AI overviews and AI search here's what's actually moving the needle. 👇

i'm building a Laravel platform that connects everything i do in my business. my obsidian vault, my tasks, my AI agents, my client projects; one operational layer. having everything talk to each other is where i see AI and agents going. not isolated tools; connected systems.

planning is the highest leverage part of the agentic engineering process. AI compressed the build step. lines of code that used to take days can take as little as 15min. but if you skip planning, you just get to the wrong answer faster.

upgraded a client project from Laravel 9 to 12 today. three major versions, swapped Nova for Filament, full test suite and manual testing. Laravel Shift + AI with a few judgment calls on my part but mostly automated in one day. done mostly in the background while working on other projects.

AI helps me leverage the things i'm good at; architecture, taste, client conversations; and speed-run the repetitive stuff i used to procrastinate on. it didn't replace my strengths. it gave me more time to use them.

outsource execution to AI. use AI to partner on decisions and thinking. use AI to help you learn. but don't outsource knowledge and understanding. that's the part that compounds.

the cost of building internal tools dropped so much that "just deal with it" stopped making sense. hit a friction point, build a tool, move on. most of them take an afternoon now. the best ops improvements aren't bought. they're built for your exact workflow.

models get replaced every few months. your system survives the upgrade. the memory files, the workflows, the conventions; they transfer to the next model and the one after that. bet on the system, not the model.

the loudest AI conversation is about which model is smartest. the quiet advantage is the person who spent six months building systems around a "good enough" model and now ships circles around everyone debating benchmarks.

Laravel has mail, notifications, queues, jobs, a package system, Eloquent, and now an AI SDK built in. AI understands the framework well. the ecosystem is mature. you can go from idea to production faster than ever. there's never been a better time to build with Laravel. Laravel forever.

i spend more time tweaking my AI system than i probably should. new agent for this. better prompt for that. one more memory file. but every hour invested pays back across every project i touch. so i keep going.

don't optimize the prompt. optimize the context. your notes, your conventions, your past decisions; all feeding into every interaction. that's where the real leverage is.

i've accumulated years of notes in my obsidian vault. every meeting, every decision, every mistake; all searchable, all linked. the compounding isn't in the notes themselves. it's in how fast i can recover context i would've lost.