You can now block users directly from comments on personal GitHub repos! https://github.blog/changelog/2026-08-13-block-users-from-comments-in-personal-repositories/ I sense that I'll be using this soon and often...

Block users from comments in personal repositories - GitHub Changelog

You can now block or unblock users directly from comments on pull requests and issues in repositories owned by personal GitHub accounts. To block or unblock a user: Open the…

github.blog

Wrote a new blog post: "Building safe servers for your PostgreSQL database" https://blog.pamelafox.org/2026/08/building-safe-mcp-servers-for-your.html That's based on my @posetteconf talk, recording here: https://www.youtube.com/watch?v=3_JPHuXgDyQ

Open Source AI Gap Map: The most solid open-source options are in the orchestration & agents category, like the many Python agent frameworks, but we need more OSS options around evals, UI, OTel, and more. https://www.aipotluck.org/map

Building with MCP in the Bay Area? Join us Sept 14 at GitHub HQ for MCP Community Connect SF! Hear from our fantastic speakers: • Drew Breunig (CMPND) • Jiquan Ngiam (MintMCP) • Sofía Sánchez-Zárate (CopilotKit) • Prathmesh Patel (MCPJam) • Harald Kirschner […] [Original post on fosstodon.org]

Just discovered Hillel Wayne's Crossover Project, where he interviews a bunch of "trad" engineers who became SWEs, to decide if SWE is really "engineering". I know we all have gut reactions, but read it! https://www.hillelwayne.com/tags/crossover-project/

GitHub launched stacked PRs today! https://github.blog/changelog/2026-07-30-stacked-pull-requests-are-now-in-public-preview/ Who's started using them already? Link me to your stack. 🥞

Stacked pull requests are now in public preview - GitHub Changelog

Stacked pull requests break large changes into small, reviewable pull requests. They’re an ordered series of pull requests that each represent focused layers of your change. With stacks, you can…

github.blog

The new MCP spec is officially out! https://blog.modelcontextprotocol.io/posts/2026-07-28/ To celebrate, my colleague and I are hosting a 5-hour livestream on September 9 on the VS Code YouTube channel. Join MCP maintainers and builders for talks about MCP […] [Original post on fosstodon.org]

Released new version of axe-core-playaxe-playwright-python for automated a11y testing of websites: https://pypi.org/project/axe-playwright-python/ Thank you to the community for their PRs keeping it up to date with latest axe-core accessibility checks.

"Do automated evals work?" https://parlance-labs.com/blog/posts/auto-evals/ Hamel and Antaripa use multiple LLM-powered tools to identify failures in app traces and compare results to labels from a domain expert. LLMs caught some errors that humans missed, but […] [Original post on fosstodon.org]

Great intro to git worktrees from @nzakas: https://humanwhocodes.com/blog/2026/07/introduction-git-worktrees/ The GitHub Copilot app uses worktrees by default, and it reallly helps to understand them, since a worktree will *not* have files from the main checkout like a gitignored .env.

A gentle introduction to Git worktrees - Human Who Codes

Git worktrees let you check out multiple branches into separate directories simultaneously, making parallel local development straightforward whether you're working solo or alongside AI coding agents.

humanwhocodes.com

Next week, my colleague and I are running a livestream series showing how to ground AI apps in real data: "Microsoft IQ Deep Dive with Python" We'll cover: * Foundry IQ (Azure AI Search): create multi-source knowledge bases with hybrid search and re-ranking * […] [Original post on fosstodon.org]

Great intro to "product quantization", a vector compression technique used by ColBERT. https://isaacflath.com/writing/QuantizationFundamentalsForMultiVectorRetrieval

Quantization for ColBERT | Isaac Flath

A thorough introduction to Quantization for Multi-Vector Search Architectures

isaacflath.com

As a life-long Photoshop addict, I love that the web is now powerful enough for basic photo editing operations. I've been using http://photopea.com, as it's free and mimics the Photoshop UI and shortcuts.

My current guilty pleasure is watching “Love is Blind: Argentina” - but I figure it’s educational, since it’s all in spanish! Argentinian spanish is a fascinating dialect, it has a lot of sound changes: https://en.wikipedia.org/wiki/Rioplatense_Spanish

Rioplatense Spanish - Wikipedia

en.wikipedia.org

my 7-yr-old wants a karaoke machine for her bday! how do karaoke machines work in the modern age? notably, we have no TV, only a projector with chromecast. any recommendations?

Super interesting paper: "Prompt Injection as Role Confusion" https://role-confusion.github.io/ They trained a linear probe to predict roles (user/tool/assistant), and then experimented with injecting role tags and just changing the writing style to mimic a role (like <think>ing).

Prompt Injection as Role Confusion

LLMs can't tell who's speaking. We show they identify roles by writing style, not tags, and exploit this with CoT Forgery, injecting fake reasoning that models mistake for their own thoughts.

role-confusion.github.io

My colleague and I are running a new livestream series: Microsoft IQ Deep Dive with Python We'll cover: * Foundry IQ (Azure AI Search): create multi-source knowledge bases with hybrid search and re-ranking * Fabric IQ: query ontologies from MS Fabric * Work IQ […] [Original post on fosstodon.org]

I love my cute lil iPhone 13 mini that fits in all my pockets, but its battery is on its last legs. What's the dantiest modern iPhone I can get?

You can now set a limit for the number of allowed open PRs for users without write access on GitHub repos: https://github.blog/changelog/2026-06-17-limit-open-pull-requests-for-users-without-write-access/

Limit open pull requests for users without write access - GitHub Changelog

Maintainers of open source repositories are dealing with an ever-growing volume of pull requests, including repeated low-quality or drive-by contributions that can slow triage and overwhelm review queues. To help…

github.blog

My Posette talk is now on YT- "An MCP for your Postgres DB" https://youtube.com/watch?v=3_JPHuXgDyQ I build a Python FastMCP server, starting with free-form SQL tools, then read-only tools, then typed query-specific tools. Slides: https://pamelafox.github.io/mcp-for-postgres-db-demo/ Code […]

Helpful post from my colleague: "What are git worktrees, and why should I use them?" Both VS Code and GitHub Copilot app have built-in support for worktrees, and they can be really useful if you get how they work […]

I just merged an update to the Azure AI Search + OpenAI RAG project that changes the default chat model to gpt-5.4-mini with reasoning_effort=low and top=5. https://github.com/Azure-Samples/azure-search-openai-demo/pull/3038 That combo wasn't best on every […] [Original post on fosstodon.org]

I tried out Pilates 6 months ago and I love it - started 1x a week and am now at 2x-3x a week. I like how it involves so much stretching and how all the different equipment configurations keeps things interesting. So many creative ways to plank!

My 6 yr old is just singing loudly to herself over and over "I'm blue, if I was green, I would die." She gets that from me. 😊

Got a message from an Amazon recruiter, which reminds me that my earliest rejection was in 2006 after an Amazon phone screen. I was asked how to set a cookie in JS, and I couldn't say exactly how. I know how now, but still don't think I've ever hand-written the code to do that.

I love keyless connections, so I'm happy that the Azure PostgreSQL team made a library specifically for Entra OAuth token refresh, compatible with psycopg3, psycopg2, and SQLAlchemy: https://pypi.org/project/azure-postgresql-auth/

azure-postgresql-auth

Microsoft Azure PostgreSQL Auth Library for Python

pypi.org

Want a token-efficient mode for VS Code Copilot? Share your thoughts on this open issue where they're exploring that: https://github.com/microsoft/vscode/issues/320470

Explore an "efficiency mode" selection for chat · Issue #320470 · microsoft/vscode

In an effort to help users optimize their AI credits, we should explore adding an "efficiency" or "low cost" mode that can be toggled on for any model. It would enforce changes like: default reason...

github.com

The Azure Postgres team open-sourced a new extension, pg_durable, for durable execution built directly into PostgreSQL, coded with a DSL and executed with Rust duroxide. https://techcommunity.microsoft.com/blog/adforpostgresql/introducing-durable-functions-in-postgresql/4526821 Plus the VS […]

Alright, slackers. This is it. Time to submit your @pybeach proposals. You know you want to! https://2026.pybeach.org/speak #PyBeach #CfP

Is there a metric like "token hungriness" or "token thriftiness", for models in agentic coding scenarios? It'd be great to figure out which models are good at minimizing unneeded tool calls and excessive codebase exploration.