We covered 10+ AI dev tools last week. Here are the top 10; 8 of them are open source.
Git for hundreds of agents, sandboxes that sleep, a voice agent you can interrupt, a red-team kit for MCP servers.
The list 🧵
#OpenSource #AIDevTools
Git for hundreds of agents, sandboxes that sleep, a voice agent you can interrupt, a red-team kit for MCP servers.
The list 🧵
#OpenSource #AIDevTools
November 8, 2026 at 10:28 PM
We covered 10+ AI dev tools last week. Here are the top 10; 8 of them are open source.
Git for hundreds of agents, sandboxes that sleep, a voice agent you can interrupt, a red-team kit for MCP servers.
The list 🧵
#OpenSource #AIDevTools
Git for hundreds of agents, sandboxes that sleep, a voice agent you can interrupt, a red-team kit for MCP servers.
The list 🧵
#OpenSource #AIDevTools
Learning point from project 1: the dev container starter had no Node, uv or Postgres option, so it was rebuilt several times mid-project. The updated template pins Node and uv, puts the agent CLIs on PATH and adds a Postgres. #aidevtools
November 4, 2026 at 11:41 PM
Learning point from project 1: the dev container starter had no Node, uv or Postgres option, so it was rebuilt several times mid-project. The updated template pins Node and uv, puts the agent CLIs on PATH and adds a Postgres. #aidevtools
Preparing for Project 2 starts with the agent kit. The starter docs, the update path for installed files and the README were reworked first, so the new project begins on an improved process, not the one project 1 outgrew. Kit 1.10 is the target. #aidevtools
November 4, 2026 at 8:12 PM
Preparing for Project 2 starts with the agent kit. The starter docs, the update path for installed files and the README were reworked first, so the new project begins on an improved process, not the one project 1 outgrew. Kit 1.10 is the target. #aidevtools
17/ Peer reviewing a traffic-monitoring platform (Kafka → Postgres → FastAPI → React). The best part is how honest the docs are. They separate “implemented” from “verified”, and say outright what’s unproven. It makes reviewing easy. #aidevtools
October 29, 2026 at 9:50 AM
17/ Peer reviewing a traffic-monitoring platform (Kafka → Postgres → FastAPI → React). The best part is how honest the docs are. They separate “implemented” from “verified”, and say outright what’s unproven. It makes reviewing easy. #aidevtools
15/ Peer reviewing an HVAC-R catalogue app (FastAPI + React + Postgres). Its OpenAPI spec is generated from the code, and CI fails if it drifts. Docs that can’t go stale are a quiet superpower. #aidevtools
October 29, 2026 at 9:25 AM
15/ Peer reviewing an HVAC-R catalogue app (FastAPI + React + Postgres). Its OpenAPI spec is generated from the code, and CI fails if it drifts. Docs that can’t go stale are a quiet superpower. #aidevtools
19/ Peer reviewing an apartment association app (Vue + FastAPI + Postgres) that does the full course loop. Alert → on-call agent → Claude Code → fix PR, all wired up with a recorded run. Most projects stop at the alert. #aidevtools
October 29, 2026 at 11:13 AM
19/ Peer reviewing an apartment association app (Vue + FastAPI + Postgres) that does the full course loop. Alert → on-call agent → Claude Code → fix PR, all wired up with a recorded run. Most projects stop at the alert. #aidevtools
20/ The same repo makes privacy a tested rule, not a promise. No names, emails or phones in the API, only masked badges like “Apt. 14 • OWNER”, and telemetry is scrubbed too. Its README even maps each course criterion to where it’s met. #aidevtools
October 29, 2026 at 11:13 AM
20/ The same repo makes privacy a tested rule, not a promise. No names, emails or phones in the API, only masked badges like “Apt. 14 • OWNER”, and telemetry is scrubbed too. Its README even maps each course criterion to where it’s met. #aidevtools
16/ Best part of the repo I reviewed: a real incident write-up. The UI loaded but the catalogue was empty, because a .dockerignore dropped the seed file and a broad except hid the error. A quiet failure is worse than a loud one. #aidevtools
October 29, 2026 at 9:25 AM
16/ Best part of the repo I reviewed: a real incident write-up. The UI loaded but the catalogue was empty, because a .dockerignore dropped the seed file and a broad except hid the error. A quiet failure is worse than a loud one. #aidevtools
18/ The same repo has an “on-call agent” that is really a fixed script: one alert, one restart, no LLM. Good automation, but not an agent that investigates and adapts. Don’t stop at a runbook, hand the alert to a coding agent with read-only access. #aidevtools
October 29, 2026 at 9:51 AM
18/ The same repo has an “on-call agent” that is really a fixed script: one alert, one restart, no LLM. Good automation, but not an agent that investigates and adapts. Don’t stop at a runbook, hand the alert to a coding agent with read-only access. #aidevtools
14/ Every PR runs the tests. A merge to main deploys to dev on AWS. Production gets the same image through a manual run that needs my approval, and I’ve rehearsed a rollback. #aidevtools
October 26, 2026 at 5:41 AM
14/ Every PR runs the tests. A merge to main deploys to dev on AWS. Production gets the same image through a manual run that needs my approval, and I’ve rehearsed a rollback. #aidevtools
11/ Supercharged velocity by integrating Kiro agentic workflows into VulnTriage. Connected AWS IAM Identity Center access and linked cloud sessions directly to GitHub. Now Kiro tackles complex backend tasks like OSV vulnerability matching and autonomously opens clean pull requests. #aidevtools
October 24, 2026 at 12:48 AM
11/ Supercharged velocity by integrating Kiro agentic workflows into VulnTriage. Connected AWS IAM Identity Center access and linked cloud sessions directly to GitHub. Now Kiro tackles complex backend tasks like OSV vulnerability matching and autonomously opens clean pull requests. #aidevtools
13/ Unit tests cover the parsers and the ranking tiers. Integration tests run against real Postgres in Docker Compose, with stored feed responses so CI never depends on live vulnerability feeds. make test runs the fast suite; one more command runs the Postgres integration tests. #aidevtools
October 24, 2026 at 4:55 AM
13/ Unit tests cover the parsers and the ranking tiers. Integration tests run against real Postgres in Docker Compose, with stored feed responses so CI never depends on live vulnerability feeds. make test runs the fast suite; one more command runs the Postgres integration tests. #aidevtools
12/ The FastAPI backend matches each dependency against OSV.dev, then ranks findings with CISA KEV and EPSS data plus asset context. The ranking is a fixed formula, not a model, so every priority can be explained. SQLite for development, Postgres in Compose. #aidevtools
OSV - Open Source Vulnerabilities
Comprehensive vulnerability database for your open source projects and dependencies.
OSV.dev
October 24, 2026 at 3:10 AM
12/ The FastAPI backend matches each dependency against OSV.dev, then ranks findings with CISA KEV and EPSS data plus asset context. The ranking is a fixed formula, not a model, so every priority can be explained. SQLite for development, Postgres in Compose. #aidevtools
10/ The API contract comes from the frontend’s services layer: an agent writes openapi.yaml from it, and a backend test fails if FastAPI’s generated document drifts from it. The frontend, n8n and the MCP server all build against the same endpoints. #aidevtools
October 22, 2026 at 9:29 AM
10/ The API contract comes from the frontend’s services layer: an agent writes openapi.yaml from it, and a backend test fails if FastAPI’s generated document drifts from it. The frontend, n8n and the MCP server all build against the same endpoints. #aidevtools
9/ Usage limits hit quickly, but I had free Claude Code cloud credits. To run web sessions safely, I mapped out repo changes first: pre-installed Docker/Postgres unblocks integration tests, while strict GitHub rulesets prevent agents pushing to main. #aidevtools
October 22, 2026 at 3:38 AM
9/ Usage limits hit quickly, but I had free Claude Code cloud credits. To run web sessions safely, I mapped out repo changes first: pre-installed Docker/Postgres unblocks integration tests, while strict GitHub rulesets prevent agents pushing to main. #aidevtools
8/ A custom setup script pins the toolchain (uv 0.13.1, Node 24.22.0), disables unneeded Corepack shims, and handles persistent storage for agent configs via Docker named volumes (/home/vscode/.claude). Clean, reproducible, and contained. #aidevtools
October 22, 2026 at 12:18 AM
8/ A custom setup script pins the toolchain (uv 0.13.1, Node 24.22.0), disables unneeded Corepack shims, and handles persistent storage for agent configs via Docker named volumes (/home/vscode/.claude). Clean, reproducible, and contained. #aidevtools
7/ Giving coding agents terminal access requires security boundaries. I created an agent sandbox using Dev Containers: SSH auth disabled (SSH_AUTH_SOCK=""), interactive git prompts suppressed, and credentials scoped via an isolated, read-only AGENT_GH_TOKEN. #aidevtools
October 21, 2026 at 7:41 AM
7/ Giving coding agents terminal access requires security boundaries. I created an agent sandbox using Dev Containers: SSH auth disabled (SSH_AUTH_SOCK=""), interactive git prompts suppressed, and credentials scoped via an isolated, read-only AGENT_GH_TOKEN. #aidevtools
5/ Two coding agents share the work: OpenCode on free models and Claude Code. Either can build, depending on quota; additional review runs on Claude Code where possible. #aidevtools
October 18, 2026 at 10:45 PM
5/ Two coding agents share the work: OpenCode on free models and Claude Code. Either can build, depending on quota; additional review runs on Claude Code where possible. #aidevtools
6/ Frontend first, as a discovery build. Lovable and Bolt both built prototypes from the spec. Bolt’s free credits went much further, but Lovable’s version matched the requirements more closely, so it went forward. Every backend call sits in one mocked services layer. #aidevtools
October 18, 2026 at 11:23 PM
6/ Frontend first, as a discovery build. Lovable and Bolt both built prototypes from the spec. Bolt’s free credits went much further, but Lovable’s version matched the requirements more closely, so it went forward. Every backend call sits in one mocked services layer. #aidevtools
4/ Work runs through a small agent team: a PM agent grooms each GitHub issue, an engineer agent implements it, and a QA agent checks it against the acceptance criteria, on a different tool from the build where possible. Protected files change only with human approval. #aidevtools
October 18, 2026 at 8:30 AM
4/ Work runs through a small agent team: a PM agent grooms each GitHub issue, an engineer agent implements it, and a QA agent checks it against the acceptance criteria, on a different tool from the build where possible. Protected files change only with human approval. #aidevtools
3/ Before touching code, I mapped out a product spec covering asset ingestion, CVE matching rules, and risk scoring. Giving the agent clear boundaries upfront stops it from over-engineering unnecessary features. #aidevtools
October 18, 2026 at 4:42 AM
3/ Before touching code, I mapped out a product spec covering asset ingestion, CVE matching rules, and risk scoring. Giving the agent clear boundaries upfront stops it from over-engineering unnecessary features. #aidevtools
2/ I’ve settled on VulnTriage: a full-stack vulnerability exposure tracker that matches asset inventories against public vulnerability data and ranks what to patch first. Clear user stories will keep the implementation disciplined. #aidevtools
October 18, 2026 at 4:10 AM
2/ I’ve settled on VulnTriage: a full-stack vulnerability exposure tracker that matches asset inventories against public vulnerability data and ranks what to patch first. Clear user stories will keep the implementation disciplined. #aidevtools
Kicking off Project 1 for #aidevtools. Building a security tool means getting the workflow right before writing any code. Spec-driven development is the best way to keep a coding assistant focused on real architecture instead of guessing.
October 18, 2026 at 2:50 AM
Kicking off Project 1 for #aidevtools. Building a security tool means getting the workflow right before writing any code. Spec-driven development is the best way to keep a coding assistant focused on real architecture instead of guessing.
OpenAI is rolling out new visual ad formats in ChatGPT alongside expanded measurement, attribution, and brand suitability tools for advertisers. The move signals how advertising may be woven more…
#ChatGPT #AD #OpenAI #AIDevTools
https://openai.com/index/new-chatgpt-ads-format-and-measurement
#ChatGPT #AD #OpenAI #AIDevTools
https://openai.com/index/new-chatgpt-ads-format-and-measurement
October 5, 2026 at 12:01 PM
OpenAI is rolling out new visual ad formats in ChatGPT alongside expanded measurement, attribution, and brand suitability tools for advertisers. The move signals how advertising may be woven more…
#ChatGPT #AD #OpenAI #AIDevTools
https://openai.com/index/new-chatgpt-ads-format-and-measurement
#ChatGPT #AD #OpenAI #AIDevTools
https://openai.com/index/new-chatgpt-ads-format-and-measurement