标签：#productivity

The first sign something was wrong wasn't a crash. It was a pattern. blockly-platform was the first real thing I built with Claude Code end to end — a Blockly-based platform for university programming exercises, driven entirely through Claude Code's Telegram channel. No editor open, no repo checked out on my machine, just a chat thread. I'd describe what I wanted, Claude Code would build it on a box I never looked at directly, and I'd judge the result by clicking around the deployed app. On March 22nd, the home page came up empty. GET /api/exercises/published was returning 403. I said so in the chat; a few messages later, Claude Code said it was fixed — the endpoint hadn't been added to Spring Security's permitAll() list. I moved on, tried the category filter. Also empty, also 403, also missing from the same permitAll() list — same file, same class of fix, different line. Then the exercise detail page. Same story, third time, same day. Three days later, the like button stopped working — root cause, again: POST /api/exercises/*/like had never been whitelisted either. Four times, one file, one recurring gap. None of these were hard bugs. Each one, in isolation, is a one-line fix a competent engineer makes without thinking twice. What bothered me, once I noticed the pattern, was that I hadn't noticed it as it happened. I had no diff to scroll through, no file to glance at and think "wait, didn't we just fix this exact class of thing twice already?" I had a chat log and a live app to poke at. The fourth fix looked, from where I sat, exactly like the first: a message telling me it was resolved. That was the moment I started to suspect the problem wasn't the model. It was that nobody — not the model, not me — had anything to look at. Why chat-only vibe coding breaks down Here's what makes that pattern more interesting than "the AI made a mistake": every one of those four fixes was correct. Claude Code read the error, found the missing permitAll() entry, added it, and move

Cet article a été publié à l'origine sur WP Admin Lab , le journal du web technique en français. WP-CLI est l'interface en ligne de commande officielle de WordPress, un outil indispensable pour tout développeur ou administrateur gérant des sites WordPress en production. Contrairement à l'interface graphique du tableau de bord, WP-CLI permet d'exécuter des opérations en masse, d'automatiser des tâches répétitives et d'intervenir sur des sites inaccessibles via le navigateur. En 2026, maîtriser WP-CLI est devenu une compétence fondamentale pour la gestion professionnelle de parcs WordPress, que ce soit pour des agences gérant des dizaines de sites ou pour des développeurs travaillant sur des environnements de staging et de production. Installation et configuration de WP-CLI en 2026 L'installation de WP-CLI s'effectue en téléchargeant le fichier Phar officiel depuis le dépôt GitHub du projet et en le rendant exécutable à l'échelle du système. Sur les serveurs Linux mutualisés ou dédiés, la commande curl -O permet de récupérer le binaire, que l'on déplace ensuite vers /usr/local/bin/wp avec les droits d'exécution appropriés. Les hébergeurs comme Kinsta ou WP Engine proposent WP-CLI préinstallé dans leurs environnements SSH, facilitant la prise en main immédiate. La vérification de l'installation avec wp -info fournit les détails de version, l'interpréteur PHP utilisé et le chemin vers le fichier de configuration wp-config.php. La configuration avancée de WP-CLI passe par le fichier wp-cli.yml placé à la racine du projet WordPress. Ce fichier YAML permet de définir l'URL du site, le chemin d'installation, les alias de serveurs distants pour les déploiements SSH, et des paramètres par défaut pour certaines commandes. Les alias SSH dans wp-cli.yml sont particulièrement puissants : ils permettent d'exécuter des commandes WP-CLI sur un serveur distant exactement comme en local, avec une syntaxe du type wp @production plugin list. Cette fonctionnalité simplifie considérableme

How a distributed, event-sourced issue tracker built with developer ergonomics in mind may have a role to play in the next generation of agentic workflows Vision drift Harness engineering has recently popularized the idea of containing architectural drift in agentic workflows. What might be missing in the discussion is a similar issue on a higher level - vision drift . By vision drift I mean that the implementation no longer drifts only from the architecture - it drifts from the original product intent. And it seems like the risk may be obscured by restricted tooling. As long as the project management tools only present a snapshot rather than a traceable story, there is an increased risk of undetected drift. Drift is detected via specification audits over time. However, while code history easily can be traversed via Git, issue tracking essentially lacks this capability. Issue trackers tend to be excellent at answering the question “what is going on right now?”, but fail at answering the question “how did our work in this area evolve last month?” or “what went on this time last year?”, or “how did we get from there to there?”. Workflow audits When I set off to build Epiq, this was not a concern on my radar. Agentic coding was something I had heard distant rumors of, and in fact I was just pursuing the ideal developer experience . This pursuit did however lead me down a path of unorthodox architecture, which in turn resulted in an issue tracker with some uncommon properties. One of these is the ability to inspect historical state by time-traveling, and replay sequences. I have not yet encountered another issue tracker with these capabilities. Initially I thought of it as a gimmick feature. Imagine the wow-factor of replaying the entire sprint in a retro, visualizing the past 2 weeks as a short movie. I thought it would help out with reflection of how much (or little) work had been accomplished. Not until I set out to do my own first fully agent-implemented feature did

Scope creep usually does not start with a dramatic disagreement. It starts with a reasonable-sounding request: "Can we also add this small thing?" For freelancers, consultants and small agencies, the hard part is often not the work itself. The hard part is that nobody has a clean shared record of what changed, what was included originally, what is now extra, and whether the timeline or price should move. The workflow I find most useful is deliberately simple. 1. Write the baseline before work starts Before the project begins, write down: the goal of the work what is included what is excluded assumptions dependencies what the client needs to provide This does not need to be legal language. Plain language is often better because both sides can understand it quickly. 2. Separate included and excluded work Many scope problems happen because "not included" was never written down. For example: Included: one landing page with copy and layout Excluded: email automation, paid ad setup, analytics dashboard, extra page variants That makes later conversations less emotional. You are not saying "no" from nowhere. You are comparing the new request to the original scope. 3. Pause before doing the extra work When a new request appears, write it down before starting. A useful change note can be short: requested change why it is needed impact on price impact on timeline what will be delivered approval status The point is not bureaucracy. The point is to stop invisible work from becoming normal. 4. Make timeline impact explicit Freelancers often talk about price but forget schedule. Even if the client accepts the extra cost, the original delivery date may no longer be realistic. A small change can still interrupt review cycles, dependencies or other client work. 5. Get written approval before starting This can be as simple as: "Confirmed. Please go ahead with this change at the updated price and timeline." The approval does not need to be fancy. It just needs to exist before the addit

TL;DR: Graph renderings of blockchain transactions carry topology signals that serialize badly into token sequences. A hub node surrounded by 47 short-lived leaf wallets looks like a table of addresses and amounts in text form — recognizable only if you already know the pattern. 📖 Reading time: ~23 min What's in this article The Problem: Blockchain Forensics Needs Vision, Not Just Text Hardware and Environment Setup on MI300X Data Pipeline: Rendering Blockchain Graphs as Training Images Fine-Tuning Loop: LoRA on 7B vs Full-Parameter on 7B ROCm-Specific Failure Modes and How to Diagnose Them Inference Serving: vLLM on ROCm for Classification Throughput Verdict: When This Setup Makes Sense and When It Doesn't The Problem: Blockchain Forensics Needs Vision, Not Just Text Graph renderings of blockchain transactions carry topology signals that serialize badly into token sequences. A hub node surrounded by 47 short-lived leaf wallets looks like a table of addresses and amounts in text form — recognizable only if you already know the pattern. Rendered as an image, that star topology is immediately visible as a structural shape. The same applies to layering patterns in mixing operations, where funds move through sequential depth levels that form visually distinct bands, and to clustering signatures where tightly-coupled address groups show dense internal edges versus sparse external ones. A vision-language model can learn to classify on those shapes directly. A text-based LLM working from a transaction list has to reconstruct the topology from raw numbers, which is possible but brittle — edge count and clustering coefficient can be computed and injected as tokens, but that's you doing the feature engineering that the vision model can learn to do itself. The reason Qwen2-VL entered this experiment rather than a GNN is mostly practical. Graph neural networks are the academically correct tool for graph classification, but they require a fixed-schema graph dataset and a trainin

On 23 June 2026, AWS shipped Lambda MicroVMs : isolated VMs you launch, suspend, resume and terminate through an API, built explicitly for "workloads that execute user- or AI-generated code." Up to 16 vCPUs, 32 GB of memory, 8 hours of runtime, a dedicated HTTPS endpoint per VM. We've been shipping that product for a while. So has E2B, so has Modal. The interesting thing about the launch isn't that AWS caught up - it's that the biggest infrastructure company in the world looked at agent sandboxes, agreed with the design, and then shipped it with one European region and a price roughly three times ours per vCPU . That's the whole post. If you want an AWS Lambda MicroVMs alternative, "can anyone else do this" isn't the question - the isolation technology is literally the same on both sides. The questions are who operates the machine, where in Europe you can put it, and what it costs to leave running. AWS wins some of these outright, and we'll say where. The short verdict Pick Lambda MicroVMs if you're already deep in AWS, need more than 4 vCPUs or 8 GB in a single sandbox, or your agent has to reach private resources inside your VPC. The IAM integration and the size of the fleet are real advantages. Pick orkestr sandboxes if you're an EU company that wants execution, snapshots and logs inside one EU legal entity, you want to pay for CPU you actually burned rather than CPU you reserved, and your sandboxes are small and numerous rather than huge. The same primitive Both products run on Firecracker. AWS says so on the docs page - "Lambda MicroVMs deliver these core capabilities through Firecracker virtualization" - and so do we. Each sandbox is a hardware-isolated VM with its own kernel and rootfs, not a container sharing the host kernel. That distinction matters exactly when an LLM is writing shell commands you haven't read yet. The lifecycle is the same too. Create, exec, read and write files, pause, resume, terminate. Here's ours: from orkestr import Sandbox with Sand

If you've run more than one AI coding agent on the same project, you already know the failure mode. You point Claude Code at /src/game and Cursor at /src/ui "just to be safe," and twenty minutes later one of them has quietly rewritten a file the other was mid-edit on. No error, no warning — just a diff that makes no sense and an afternoon spent figuring out which agent ate whose work. The agents aren't the problem. The problem is that multiple AI coding agents on the same codebase have no shared notion of "someone else is touching this file right now." Each one acts as if it's alone, and that assumption breaks the moment you run two, three, or four in parallel — exactly when a solo builder or vibe-coder would want to, to ship faster. I built Bothread to fix this. It's free, open-source, and runs entirely on your own machine. Why AI Coding Agents Overwrite Each Other's Files The core issue is coordination, not intelligence. One agent working alone is usually fine. Trouble starts when a second agent, unaware of the first, opens that same file and writes its own version on top. Whoever saves last wins, silently — no lock, no claim, no message saying "I'm in physics.js , give me five minutes." Multiply that by however many agents you're running and you get the pattern anyone doing multi-agent AI coding eventually hits: duplicated work, clobbered edits, and a human reconstructing what happened after the fact instead of watching it happen. Bothread's answer: give the agents a shared room, over MCP (Model Context Protocol) , where "who's working on what" is a fact everyone can see and act on — not something you guess at after a merge conflict. What Bothread Actually Does Bothread is a small local server (no cloud, no accounts) that any MCP-compatible agent can join as a participant in a shared room: Claim files before editing — a claim on a file someone else already holds gets denied and shown, instead of silently overwritten. Talk in a live thread , share a task board and

I had a tidy little helper that computed a thinking budget based on input size. Something like "give the model 30% of the context as thinking room." It worked great on Opus 4.5. Then I tried to point it at Opus 4.8 and got a 400. The whole concept I had built around is gone in the current models. Here is what replaced it and how I migrated. What broke The old pattern looked like this: // Opus 4.5 and earlier const response = await client . messages . create ({ model : " claude-opus-4-5 " , max_tokens : 16000 , thinking : { type : " enabled " , budget_tokens : 8000 }, messages , }); On Opus 4.7, 4.8, and Fable 5, thinking: { type: "enabled", budget_tokens: N } returns a 400. The fixed token budget is dead. The replacement is adaptive thinking, where the model decides how much to think, plus an effort knob that controls overall token spend. // Opus 4.8 const response = await client . messages . create ({ model : " claude-opus-4-8 " , max_tokens : 16000 , thinking : { type : " adaptive " }, output_config : { effort : " high " }, // low | medium | high | xhigh | max messages , }); Why this is actually better (after I got over it) My old budget code was a guess dressed up as a calculation. I had no real basis for "30% of context." I picked it because it felt reasonable and the outputs looked fine. Adaptive thinking moves that decision to the model, which sees the actual problem. The mental model shift: budget_tokens controlled how much the model could think. effort controls how much it thinks and acts . They are not the same axis, so there is no clean 1:1 mapping. I stopped trying to translate "8000 tokens" into an effort level and instead picked based on the workload. How I chose effort levels After running my own evals, here is where I landed: Workload Effort Notes Classification, routing low Fast, scoped, not intelligence-sensitive Most app traffic medium to high The balance point Coding and agentic loops xhigh Best for these; it is the Claude Code default Correctness

Introduction So imagine you are focused on your cappuccino-frappuccino doing something very important on you win laptop and then have a cringe attack due to the unknown Phone Link notification : Complete linking devices Your PC and mobile device are almost linked. Click here to continue linking devices. via Phone Link Then you switch off bluetooth, wifi, laptop - and you are right. What to do next ? Basic checks Settings -> Bluetooth & devices -> Mobile devices Settings -> Accounts -> Email & accounts Inspect recent notifications in Event Viewer eventvwr.msc Applications and Services Logs └ Microsoft └ Windows └ Notifications Applications and Services Logs └ Microsoft └ Windows └ Shell-Core Digital forensics Windows stores toast notifications in a local database, hence you need to install sqlite Get-ChildItem " $ env : LOCALAPPDATA \Microsoft\Windows\Notifications" output : Directory: C:\Users\$ USERNAME \A ppData \L ocal \M icrosoft \W indows \N otifications Mode LastWriteTime Length Name ---- ------------- ------ ---- d----- 1/1/2026 0:00 AM wpnidm -a---- 1/1/2026 0:00 AM 1000000 wpndatabase.db -a---- 1/1/2026 0:00 AM 10000 wpndatabase.db-shm -a---- 1/1/2026 0:00 AM 1000000 wpndatabase.db-wal wpndatabase.db is a SQLite database. connect to the database : sqlite3 " $env :LOCALAPPDATA \M icrosoft \W indows \N otifications \w pndatabase.db" query the Notification table . headers on . mode column SELECT Notification . Id , Notification . HandlerId , Notification . Type , Notification . ArrivalTime , Notification . Payload FROM Notification LIMIT 20 ; Then you will have something like : Id: [REDACTED] HandlerId: [REDACTED] Type: toast ArrivalTime: [REDACTED] Payload: <?xml version="1.0"?> <toast activationType= "protocol" launch= "ms-phone:fre/?cid=[REDACTED]&ref=FreIncompleteToast&reason=IncompleteNotificationsToast" > <visual> <binding template= "ToastGeneric" > <text hint-maxLines= "1" > Complete linking devices </text> <text> Your PC and mobile device are almost li

Everyone can generate a website now. Type a prompt, get a decent page — that part is a commodity. The question nobody's answering is what happens on day 2 : the leads start arriving, a line of copy needs a tweak, someone asks for a section you forgot. That's when a website stops being a design project and becomes a thing you have to run — and where most tools hand you yet another dashboard to log into and dread. Sitelas makes a different bet. Because a Sitelas site lives inside Claude through an MCP connector, the same chat that built the site also runs it . You don't open an admin panel to see who filled out your form, write back, or change the page. You just ask. Here's what "running your site from a chat" actually looks like. First, the 30-second why Claude connects to outside tools through MCP connectors — you already use the ones for Gmail, Calendar, and Drive. Sitelas has one too. Add it once (in claude.ai: Customize → Connectors → Add custom connector , and paste https://sitelas.com/api/mcp ), and Claude can do things with your site, not just talk about it: publish it, read its submissions, restyle it, add a section. Your site becomes an automation endpoint sitting next to your other connectors — the thing a Webflow or Squarespace site can't be. New here? Start with How to Build a Website From a Claude Chat . "Did anyone fill out my form today?" That single question is the whole idea. You ask; Claude reads your site's submissions, surfaces the new lead — Maya, a bakery owner — and drafts a warm reply in your voice. One message, no tabs. It works because every form on a Sitelas site captures submissions to your inbox automatically — no integration required. You can open that inbox in the dashboard any time: …but running your site from a chat means you rarely need to. Claude reads those same submissions straight from your site, so "who wrote in today, and what do they want?" is answered in the thread you're already in — not in a panel you have to remember to ch

My CLAUDE.md has one line near the bottom that I wrote months ago and mostly forgot about until I started actually paying attention to what it does: ## Important Note after your work done codex will review what you done. Terse, no punctuation, clearly typed in a hurry. But it's a real instruction that fires on every session in this repo: I finish a change, and a second model reviews it before I consider the work done. I added it half as an experiment. A few months in, it's changed how I work more than almost anything else in the setup, and not in the way I expected. I thought it would catch bugs. Mostly it doesn't, not directly. What it actually does is force a triage decision on every single piece of feedback, and getting that triage wrong is where all the pain lives. The three buckets Early on I treated every review comment the same way: read it, do it. That lasted about a week before I was silently making changes I didn't agree with because a second AI suggested them, and separately burning a stupid amount of time re-litigating comments that were just wrong or out of scope. What actually works is sorting every comment into one of three buckets before touching code: Fix it, no discussion. The comment is unambiguous, low-risk, and doesn't touch anything architecturally significant. Just do it and move on. Ask first. The comment is ambiguous, or it touches something that would require a real judgment call, or the "fix" would be a bigger refactor than the comment implies. Stop and get a human decision before acting. Skip silently. The comment is a duplicate of something already handled, or genuinely doesn't apply. Don't reply just to say "not doing this," don't leave a comment thread as evidence of having read it. Silence is the correct response to a non-issue. The failure mode I kept falling into before I had these buckets explicitly was collapsing 2 into 1: treating "ambiguous" as "just pick an interpretation and go." That's the actual source of review fatigue, not

Hello, I'm Maneshwar. I'm building git-lrc, a Micro AI code reviewer that runs on every commit. It is...

Why this matters I’ve worked on a few frontend projects where code reviews turned into style debates—tabs vs spaces, semicolons, quote styles… you name it. It slows everything down and adds zero value. At some point, I realized this shouldn’t even be a discussion. So now, whenever I start a project, I set up Prettier + ESLint + Husky on day one. No debates. No manual fixes. No messy PRs. This post is exactly how I do it. 🧰 What each tool actually does Prettier → formats your code automatically ESLint → catches bad patterns & enforces rules Husky → runs checks before commits (so no one skips them) Together → clean, consistent code without thinking ⚙️ Step 1 — Install dependencies npm install -D prettier eslint husky lint-staged 🎯 Step 2 — Setup Prettier Create: prettier.config.js module . exports = { semi : true , singleQuote : true , trailingComma : ' all ' , tabWidth : 2 , }; Create: .prettierignore node_modules dist build 🔍 Step 3 — Setup ESLint Initialize: npx eslint --init Then tweak your config: .eslintrc.js module . exports = { extends : [ ' eslint:recommended ' , ' plugin:react/recommended ' , ' prettier ' ], rules : { ' no-unused-vars ' : ' warn ' , ' react/react-in-jsx-scope ' : ' off ' , }, }; 👉 Important: "prettier" disables ESLint rules that conflict with Prettier. 🔗 Step 4 — Connect ESLint + Prettier Install: npm install -D eslint-config-prettier That’s it. Now ESLint won’t fight Prettier. 🐶 Step 5 — Setup Husky Initialize Husky: npx husky init Add pre-commit hook: npx husky add .husky/pre-commit "npx lint-staged" 🚀 Step 6 — Setup lint-staged Add to package.json : "lint-staged" : { "*.{js,jsx,ts,tsx}" : [ "eslint --fix" , "prettier --write" ] } 💡 What happens now? Every time you commit: ESLint checks your code Prettier formats it Only clean code gets committed No more: “fix formatting” PR comments broken lint rules in main branch inconsistent code styles 🧠 Real impact (from experience) After adding this to a team project: PR noise dropped a lot reviews

I'll Be Honest: The Internet Already Has Translators I know. Language translation isn't a new idea. There are already huge translation platforms out there. So when I started working on a translator for my tools website, I wasn't thinking: "I'm going to reinvent translation." Not at all. My thought was much simpler: "Can I make quick translation feel less distracting?" My Frustration Was Actually Pretty Simple Sometimes I just need to translate text. That's it. I don't want to: Create an account Open five different menus Break a long text into tiny pieces Jump between multiple tools I want to paste the text... Choose a language... And get the translation. So I Built My Own Version 👉 https://allinonetools.net/language-translator/ The tool currently supports 200+ languages and language variations . You can: Detect the source language Select the target language Translate long text Upload text Use voice input Listen to the result Copy or share the translation And I wanted to keep the text experience simple without forcing users into tiny input limits. Just: Enter → Choose Language → Translate 200+ Languages Sounded Simple Until I Saw the List English. Hindi. Gujarati. Spanish. Arabic. These are the languages most people immediately think about. But then I started going through the full language list. Abkhaz. Acholi. Afar. Alur. Aymara. Baluchi. And many more. Honestly... I hadn't even heard of some of them before building this. That was probably my biggest learning moment. I Realized How Small My Own View of the Internet Was As a developer, it's easy to build around the languages we personally know. For me, seeing English, Hindi, and Gujarati feels normal. But the internet is much bigger than my own experience. Someone somewhere may be trying to understand a sentence in a language I've never even heard spoken. That changed how I looked at this tool. The Hard Part Wasn't Adding a Dropdown A dropdown with 200+ options looks impressive. But that's not the real problem. The

The error looks like a typo in the remote URL. Usually it isn't. On a machine with more than one GitHub account signed in, this message is GitHub's way of saying wrong identity, not wrong address. The symptom A repo clone that has worked for months suddenly can't fetch or pull. The remote URL hasn't changed. The repo hasn't been renamed or deleted; you can open it in the browser just fine. Yet the command line insists otherwise: $ git pull remote: Repository not found. fatal: repository 'https://github.com/<org> /<repo>.git/ ' not found Why GitHub's error is misleading For a private repository, GitHub won't confirm or deny that the repo exists to a caller who isn't authorized to see it. Confirming would leak information about private repos to anyone probing URLs. So instead of a clear 403 Forbidden , an unauthorized request gets treated the same as a repo that truly doesn't exist: a 404 , which git renders as Repository not found . "Repository not found" on a private repo almost always means the credential attached to this request can't see it. It's rarely a wrong URL. The usual cause: two accounts, one keychain This shows up most on machines used for both personal and organization-owned work: a personal GitHub account for side projects, and a separate account (or SSO identity) that actually holds access to the org's private repos. Credential helpers cache one token per host. If the cached token belongs to the personal account, every git operation silently authenticates as that account, including ones against the org repo it has no rights to. personal-account --(switch)--> org-account Active, no repo access Has repo access Diagnose it First, confirm the remote itself is fine. $ git remote -v If the URL opens in a browser while logged into the right account, the remote isn't the problem. Next, check which credential is actually cached. On macOS with the default helper: $ git credential-osxkeychain get <<< $'protocol=https \n host=github.com' username=personal-account

Five litmus tests for “this will raise your intelligence” claims A pocket BS detector for brain-training ads, LinkedIn gurus, and your own wishful thinking. 1. Which dial moved? Intelligence talk smuggles four dials into one word: Dial Rough meaning g / fluid ability Harder to move; overclaimed constantly Knowledge & skill Moves with practice and education Acute sharpness Sleep, illness, mood, substances Long-horizon brain health Aging, disease risk, lifestyle If the claim does not say which dial, it is marketing soup. 2. What was the control? “People got better” is almost worthless. Better than last week of the same game is practice. Better than an active control that also gets attention, novelty, and expectation is interesting. Lumosity-style lawsuits exist because companies sold soup as medicine. 3. Near or far? Near transfer: you got better at this and close cousins. Far transfer: the effect jumped to something distant (school grades, matrix reasoning, life outcomes). Far transfer is scarce. Second-order metas on cognitive training keep finding near yes, far ≈ no once bias is handled. That is not cynicism. It is how human learning usually works. 4. Who was the sample? A processing-speed protocol that helps older adults does not automatically mint IQ points for a 24-year-old optimization bro. Deficient populations respond differently than well-nourished ones. Age and baseline matter more than branding. 5. Is the score flattering the seller? If the outcome is “our app score,” the app got better at measuring app use. Prefer outcomes that hurt to fake: standardized batteries against active controls, academic scores, carefully logged real-world performance. What I built instead of another vanity mini-game IntelligenceMax is a live reasoning gym: frontier models write fresh distinction-style items at your edge, and scoring is transparent / IRT-style. That is deliberate practice under honest difficulty, not a clinical IQ battery and not a promise that general intellige

My 2nd article this week, I'm supposed to keep it to just once per week but whateverrr, I've had this...

Most teams test email by not testing it. The send path gets a mock — expect(transport.send).toHaveBeenCalledWith(...) — and everyone agrees that's "good enough." The receive path gets skipped entirely, because there's no honest way to assert on a real inbox from a test runner. So the one part of your system that talks to the outside world over an unreliable, asynchronous, third-party channel is the part with the least coverage. That's backwards. The reason email is hard to test isn't the sending. It's the asserting . You can fire POST /messages/send all day, but to prove the message actually left, rendered correctly, and arrived with the body you expected, you need a real mailbox you control — one you can read programmatically and throw away when the run finishes. Shared Gmail test accounts almost get you there, but they bring OAuth on the runner, catch-all races between parallel workers, and a 90-day token that expires the night before a release. This post is about a different fixture: a disposable Agent Account created at the start of a CI run and deleted at the end. You mint a real mailbox per run (or per test), point your application at it, send and receive real mail, assert on the actual message body, and tear the whole thing down. No OAuth. No shared inbox. No leftover state. What an Agent Account gives you here An Agent Account is just a Nylas grant with a grant_id . That's the whole trick, and it's worth saying plainly because it's what makes this pattern cheap: an Agent Account works with every grant-scoped endpoint you already know — Messages, Drafts, Threads, Folders, Attachments, Webhooks. There's nothing new to learn on the data plane . If you've ever called GET /v3/grants/{grant_id}/messages , you already know how to read a test inbox. The difference from a normal grant is provisioning. A regular grant needs a real human to complete an OAuth flow. An Agent Account is created with a single API call — no OAuth screen, no refresh token, no human. It's a m

Most "AI email agent" demos quietly assume the agent answers everything. Point a model at the inbox, generate a reply, send it, repeat. That's a fine loop right up until the model hits a message it shouldn't touch — an angry customer, a legal question, a refund the agent has no authority to approve — and confidently fires off a reply anyway. The expensive failures in agent email aren't the threads the agent gets wrong. They're the threads the agent answers at all when it should have stepped back. So let's build the part that steps back. Not the classifier that decides a message is risky — that's triage , a separate problem. This is the handoff : once something flags a thread as "needs a human," how do you actually pull the whole conversation out of the agent's reach, park it where a person can find it, and make sure the agent keeps its hands off until that person clears it? I work on the Nylas CLI, so the terminal commands below are the exact ones I reach for when I wire up an escalation path. Every operation gets the two-angle tour: the raw curl call and the nylas command that does the same thing. What the handoff actually needs An Agent Account is, underneath, just a Nylas grant with a grant_id . That's the spine of everything here, and it's worth sitting with: there is nothing new to learn on the data plane. The same grant-scoped endpoints you already use — Messages, Threads, Folders, Drafts — work against this grant exactly the way they work against any Gmail or Microsoft grant you got through OAuth. So the escalation path isn't some special agent feature. It's three plain operations you already half-know: A place to put escalated threads. A custom folder — call it Needs human — that lives alongside the six system folders every Agent Account ships with ( inbox , sent , drafts , trash , junk , archive ). A way to move the whole thread there. Not one message — the thread . A reply is just the latest message in a conversation; a reviewer needs the full chain. A way

How I Built CodeGraph: A Living Knowledge Graph That Tells You What Breaks Before You Break It Built for HACKHAZARDS '26 — powered by Neo4j AuraDB, tree-sitter, Groq LLaMA, and Next.js The Problem That Frustrated Me Every developer knows this feeling. You join a new codebase. There are 50,000 lines of code. Your manager says "just fix this small bug in the authentication module." You make the change. You push. And suddenly three completely unrelated features are broken — a payment flow, a notification system, and a dashboard widget you've never even looked at. You spend the next four hours tracing function calls manually, reading code you've never seen, trying to understand why changing one function in auth.py broke something in notifications.py on the other side of the codebase. This is not a rare experience. According to JetBrains' developer survey, engineers spend 58% of their time reading and understanding code — not writing it. One wrong change in a large codebase can cost hours of debugging, failed deployments, and frustrated users. I built CodeGraph to solve this. Not with another AI chatbot that guesses at your code. With a real, queryable knowledge graph that actually understands how your codebase is connected. What CodeGraph Does CodeGraph takes any public GitHub repository URL and within seconds: Parses every function in the codebase using tree-sitter Maps every call relationship between functions as a directed graph Stores everything in Neo4j AuraDB as a live knowledge graph Lets you ask questions in plain English — answered by AI grounded in real graph data The result: paste a GitHub URL, see your entire codebase as an interactive graph, click any function, and instantly know what breaks if you change it. The Tech Stack Here's what I used and why each choice mattered: Backend: Python + FastAPI (REST API server) Neo4j AuraDB (graph database — the core of everything) tree-sitter (AST parser for Python, JS, TS, TSX) Groq API with LLaMA 3.3 70B (free-tier L