Why I Built a Notch App for My AI Coding Agents
Five agents. Twelve tabs. One is waiting. Which one? I run Claude Code, Codex and Gemini CLI at the same time.

Five agents. Twelve tabs. One is waiting. Which one? I run Claude Code, Codex and Gemini CLI at the same time.

Running a VPS is easy. Keeping an eye on it is another story. When you're deploying your own projects, SaaS apps, APIs, or websites, it's easy to forget what's happening on the server once everything is up and running. CPU usage, RAM, disk space, network activity, uptime, GPU information, machine details, website status... You don't always need a huge observability stack.
Public tenders are published in several places with different fields. If you track them for a product, a bid team or an agent, you end up writing a parser per source. TenderFeed returns one record shape for TED (the EU's journal) and the UK's Find a Tender service; SAM.gov can be added with your own key. curl "https://api.ambolt.dev/v1/tender-feed?keyword=software&country=DEU&free=1" The answer starts with what each source did, then the notices: { "publishedSince": "2026-09-26", "sources": [ { "source": "ted", "ok": true, "totalMatches": 285, "returned": 20 }, { "source": "uk-fts", "ok": true,
Welcome to another article in Batch Processing. We have covered Object Stores and Distributed File Systems in our previous article. It is recommended to read that before this one. In this read, we will cover distributed job orchestration and a couple of Batch Processing frameworks. Fortunately, my current job involves ownership of a huge amount of distributed job orchestration.

Two artifacts appear in every API integration: a curl command someone pasted into Slack, and a Postman collection exported from someone's account. Both are presented as "the API docs." Neither is documentation — but both are salvageable as raw material for an OpenAPI spec, as long as you know what each conversion preserves and what it silently destroys. Take a realistic capture: curl -X POST 'https://api.example.com/v1/projects?dry_run=false' \ -H 'authorization: Bearer eyJhbGciOi...' \ -H 'content-type: application/json' \ -H 'idempotency-key: 7b32f1' \ --data '{"name":"Apollo","plan":"TEAM",
OpenAPI 3.2 is a consolidation release, which is exactly what a contract format needed. The 3.0-to-3.1 jump did the disruptive work — full alignment with JSON Schema draft 2020-12, top-level webhooks, and a sane null model — and 3.2 tightened the edges. If your documents are still 3.0.3, the migration is small in mechanical terms and large in correctness, especially once you start describing streams and feeding specs to code generators and MCP servers. This is a practitioner's summary: what actually changes in your YAML, what breaks in old tooling, and how to document server-sent events proper
Article URL: https://liao.gg/blog/agents-dont-need-memory Comments URL: https://news.ycombinator.com/item?id=49945933 Points: 3 # Comments: 0
Article URL: https://www.ycombinator.com/companies/retailready/jobs/bFcgIe4-implementations Comments URL: https://news.ycombinator.com/item?id=49945904 Points: 0 # Comments: 0
Article URL: https://github.com/giga-james/eventmaxxer Comments URL: https://news.ycombinator.com/item?id=49945866 Points: 4 # Comments: 4
Too few tokens, too many spare quotas I do most of my development on a Claude Code subscription. Right now Claude Fable 5.1 and Opus 5.5 work extremely well, and I would simply call them the state of the art. The only problem is that I keep running out of tokens. At the same time, I have quota for other models lying around. I got some free quota for Xiaomi's MiMo a while ago, cheap deals for other Chinese models show up from time to time, and a friend handed me a GLM subscription they no longer use.
This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend I built this for my flatmates. The arguments in our flat are never about anything big. Dishes sitting in the sink. Someone's milk or snacks disappearing. Whose turn it is to clean, how late is too late to be loud, and a guest who stays one night longer than anyone agreed to.

Every team evaluating Model Context Protocol (MCP) servers reaches the same question: do we hand-maintain a second description of our API for agents, or do we generate it from the contract we already have? If you keep an OpenAPI document, the second answer is the sane one. An OpenAPI spec is almost literally a tool catalog already — each operation has a name, inputs, outputs, and a transport. Converting it to MCP is a mapping problem, not a modeling problem. This walkthrough covers what the mapping actually is, where generic converters fail on real specs, and how to verify the result with a re
Two days ago our little free-tools site had a perfectly good dice roller. Then it got an upgrade with delusions of grandeur. Dice Atelier is a 3D dice roller for D&D night: twenty themed tables — an astral observatory, a clockwork library, a neon ronin bar — each with dice styled to match. Mixed pools, dice notation (try 4d6+2), advantage and disadvantage, up to 100 dice at once. It's free, no sign-up: https://penguinwearingahat.com/tools/dice-atelier/ No human wrote the app.
npm create scaffold-hbar and a few minutes later you have an AI-agent passport live on Hedera testnet: an ERC-721 passport NFT, an append-only event log, snapshot attestations, a Hono API, and a tiny UI — tests green out of the box, every tx verifiable on Hashscan. We built this for the Hedera Scaffold-HBAR Template Bounty. Here's the what and the why. New-blockchain onboarding has a predictable curve: read the docs, fight the account model, fight the keys, and three hours later you still haven't deployed a contract. Hedera is worth the climb — predictable fees, fast finality — but the climb i

I'm building DevPilot AI, an open-source AI software engineering assistant designed to help developers understand, modify, test, and work with real codebases. The goal isn't just to build another chatbot. I want DevPilot to behave more like a software engineering assistant that can reason about a repository and use tools when necessary. What I'm building The current architecture has three main pieces: React — frontend For the AI layer, I'm experimenting with open-weight models running locally, including Qwen2.5-Coder through Ollama. This is important to me because I want the project to remain
What is CH-Ops CH-Ops is a browser-based operations platform built for managing ClickHouse® deployments. Instead of relying entirely on the command line or HTTP APIs, it provides a unified interface for executing SQL queries, monitoring clusters, managing users, backups, alerts, dashboards, and much more, all from a single web application. One of its useful features is Alert Rules, which lets you turn a SQL query into a scheduled check and notify you when a defined condition is met. In this walkthrough, we'll use Email (SMTP) as the notification channel and create a simple scheduled alert from
What I Built Existing online quiz tools either require a constant internet connection, distract with notifications, or force students to upload study notes to third-party cloud servers. Circuit Sathi solves this by running completely offline directly in the browser. A student simply pastes their technical lecture notes, selects the number of questions, and instantly receives tailored MCQs with explanations. Demo Initial Interface: Pasting Technical Notes: Generated Practice MCQs & Real-time Scoring: (Optional video demo link if recorded) Code {https://github.com/Hitman-886/-Circuit-Sathi-.git}
TL;DR The Midjourney vs Flux fight is decided by intent, not quality. Midjourney (V8.1, Flux (FLUX.2, pay-per-image from $0.014 to $0.07) Midjourney runs as a web app (Discord still works) with V8.1 as the default model since June 2026 — V7 shipped in April 2025, V8 in March 2026, V8.1 in April 2026 with native 2K output and roughly 3x faster/cheaper HD generation. Personalization profiles, --cref/--sref references, and an HD video model are built in. There is still no official API and no free tier. Flux is Black Forest Labs' model family, now at FLUX.2: [klein] 4B/9B (sub-second, open weights
Two documents appeared on 8 September. An arXiv paper measures what happens when a coding agent reads a poisoned repository under four permission models. A Google threat intelligence report describes malware that, inside GitHub Actions, reads the runner's OIDC token out of memory and publishes packages with valid SLSA Build Level 3 attestations. Together they say something uncomfortable about how most teams give agents permissions today, which is by writing sentences. My position: in a pipeline, an agent's permissions are whatever the runner job holds.
TL;DR The best Jasper alternatives by price are Copy.ai ($29/mo for 5 seats, unlimited Rytr ($9/mo); the best by output quality are Claude and ChatGPT Writesonic ($79–99/mo) or Surfer (from €49/mo). Jasper at $69/mo for one seat still wins on brand voice — but Jasper restructured in 2025 and the sticker got starker: Pro is $69/mo monthly, $59/mo , with no free tier (7-day trial) and a custom-priced Pricing as of October 2026 (official pricing pages): Alternative Entry price (Oct 2026) Free tier Best for Copy.ai Chat $29/mo ($24 annual) — 5 seats Yes, 2,000 words/mo Cheapest team seats, unlimit