the real bradyg/aster. works near seattle on agentic stuff. spins drum'n'bass. produces some, too. breaks stuff for fun. speaks on occasion. inventor of squad.
Joined October 2024
- Tweets234
- Following68
- Followers99
- Likes216
brady gaster retweeted
Waza started with skills. v0.39.0 goes further:
- Workflow evals
- Dashboard Lab (--lab)
- Preflight, evidence, judging & fixes
Loving the feedback! What are you testing? What's missing? Your reports & contributions shape what's next for me and the squad to deliver?
microsoft.github.io/waza/rel…
brady gaster retweeted
How do you _guarantee_ your agent's code is correct? You prove it.
Introducing Proof Driven Development Kit (PDD Kit) from @MSFTResearch.
PDD Kit helps turn your plan into a formal spec, then proves the agent's code meets that spec for every input. No vibe verification. PDD Kit is integrated with @GitHubNext's Chopin, so you can plan as a team.
PDD Kit takes an app's login flow from intent to a verified pull request:
It discusses your code, infers your intent as it goes, and proposes a formal spec in Verus, a proof language for Rust.
Once you approve the spec, the Verus verifier automatically proves the code meets it.
Where proof isn't possible, it falls back to empirical verification with tests and computer-use agents (CUAs).
It works on Rust today. No new language, no change in how you produce the underlying code, making formal methods more accessible.
The goal is a verified software factory, where engineers review specifications instead of code.
Come see it in action at GitHub Universe, Oct 28–29. Swing by the Speakeasy and ask for GitHub Next!
brady gaster retweeted
hot take alert but. me seeing all of yall rage out like, this would actually be a very low drama functional morning for me as a teen??? check your privilege mfs🤣😭
Does it bother anyone else how the members of this family communicate (or don't communicate) with each other? Negativity and entitlement just ooze from this ad everywhere.
- The first thing is the brother and sister talking rudely to each other "FINE!"
- Mom's irritated at someone for adjusting a thermostat
- "You promised French toast, bruh," ... "I'm not your bruh, bruh." ... "Fine"
- The sister sees her brother drinking milk from the jug and says, "Ugh, gross."
- She tells him French toast isn't French, and he fires back, "Yeah right, then why is it called French? Genius." She rolls her eyes and says, "Fine, live in ignorance."
- Grandpa calls and Mom shoves him over to Dad who then cuts him off with "we're a bit busy now." The grandparents go straight to insulting his beard: "We don't like the beard. Never did… He looks like that actor we hate."
- Dad answers with sarcasm ("Good pep talk") and hangs up on them.
- Brother tries to impersonate his sister and text her boyfriend.
- Mom asks if anyone else is cold, and the Dad (who we learn adjusted the thermostat) lies and says "Don't know what you're talking about."
- Mom catches Dad turning down the thermostat: "I knew it!" instead of communicating...
- The older sister barks, "Mom, breathe."
- Dad asks who Asa is, and the daughter dodges: "He's a… person."
- The whole family crowds around the screen to spy on the older sister at the front door
- Mom barks at Dad, "Bruno, we have to go now, like, now, now!"
- Dad doesn't remember the ice skating lessons (or she never told him) and when he finds out goes straight to "Hey hon… could we talk?" like he's trying to get out of it.
I mean, it demonstrates the capabilities really well. It's cool. But I feel like they could have done this from a positive, functional framing instead.
brady gaster retweeted
Nos vemos en CDMX ❤️
📢 Andrea Griffiths: Senior Developer Advocate en GitHub, con más de una década en herramientas para developers. Hace accesibles temas que van desde DevOps hasta el desarrollo asistido por IA.
@acolombiadev
brady gaster retweeted
Here's our demo from the gov.uk hackathon!
A self-service agent that helps people deal with "single justice procedures" (small crimes like speeding). Currently done manually by phone, could easily be an always-on agent
Mad trailer by @yungbose with @HyperFrames_
🥈@hturan,@Mappletons, Mevan Babakar @meandvan, @yellowcommandr & @yungbose built an AI agent that helps people who never got their court notice find their case, prove who they are and act on it.
brady gaster retweeted
Support for agents.md in Visual Studio is coming soon. It works in our internal builds and will be in the next feature update to the public Insiders build.
brady gaster retweeted
Coming up this week on Coffee & OSS I will be chatting with @1kevgriff about all sorts of #tech and #opensource topics. Streaming live on #Twitch this Tuesday August 4th at 11:00 AM PDT. Come say hello and join the conversation.
coffeeandopensource.com/sche…
brady gaster retweeted
Bro, this is cool 🤓 gitfut.com/IEvangelist/vs/da… #GitFut
brady gaster retweeted
CCR (Copilot Code Review) is on 🔥 within enterprises. Recall on Medium is now pushing 72%.
github.blog/changelog/2026-0…
brady gaster retweeted
🗣️ Calling all open source maintainers!
You can automate things like issue triage, release notes drafting, and PR auditing all inside the GitHub Copilot app!
Here's a video talking all about these automations 👇
brady gaster retweeted
The Copilot App makes the most sense when you learn the loop, not just the UI. Pick up work, steer agents, inspect changes, land the PR, and keep project state visible.
GitHub is running Let’s Learn sessions in multiple languages: gh.io/letslearn
Build smarter with the new GitHub Copilot app. 💡
Join our Let's Learn virtual sessions and find out how this all-in-one control center can help you optimize your workflow with agent-driven development.
Sign up for your session in English, Korean, Chinese, Japanese, Spanish, or Portuguese. 🗓️
gh.io/letslearn
brady gaster retweeted
If you're on windows and you want to uplevel your agent game, go to the Windows store and install the GitHub Copilot App now!!
brady gaster retweeted
BYOK is now live in the GitHub Copilot App!
Works with @ollama, foundry, and any OAI completions or Anthropic compatible messages endpoint. Give it a try today!
brady gaster retweeted
Has been a blast to collaborate with Paul on bringing impeccable into the GitHub Copilot App as a builtin.
If you’ve been struggling to get your agents to build world class UX that doesn’t feel AI generated, this is your answer!
THE NEUE CRAFT: Impeccable impeccable.style now ships inside the new GitHub Copilot app just in time for Figma #config and with generative design fully top-of-mind. Open gh.io/app and the design skill is there for you to up your level of craft, impeccably.
When I first used Impeccable it took me back twice.
- The 1980s, when PostScript arrived as a clean grammar that could describe and render any graphic you could imagine.
- The 1990s, when Kai's Power Tools showed what computation could do to an image, long before AI ✨.
Impeccable sits directly in that line. The right primitives, plus the right generative outcomes. A good tool hands you an expert instrument and leaves the refining to you. Paul's showed up just when we needed the clarifying PostScript and KPT moments again 👏 —JM HT 🧑🍳 @mariorod1 🧑🍳 @_Evan_Boyle
---
Read the post by @illscience with founder @pbakaus xitter.cf/a16z/status/2069138261…
Try out the new GitHub Copilot app with Impeccable ready-to-go gh.io/app
ALT github cat with impeccable in katakana
brady gaster retweeted
Try it out
I put together a practical sample for a thing I keep wanting as a builder:
an agent that automates my Microsoft 365 workflow, but lives in a repo, debugs locally, deploys with IaC, and runs serverlessly.
The sample is an M365 Inbox Agent on Azure Functions' Serverless Agents feature:
- urgent VIP mail -> Teams (triggered on new emails)
- action-required mail -> grounded draft replies
- daily briefing -> inbox (triggered every morning)
- rules -> markdown
- guardrails -> code, python
- auth -> Entra / managed identity
- deployment -> azd + Bicep
- local dev -> sample inbox data + dry run
This is for the custom edge cases a power user and builder hits, connected to many more systems and data sources.
These are YOUR Frontier Agents, for the last-mile stuff every team has:
- "Keep it in a repo so I can review changes"
- "Let me debug the thing before it touches real mail"
- "Reach an internal API or database that isn't in any connector catalog"
- "Pin my own model, prompt, and dependency versions, and roll back a bad one"
- "Run in my own Azure subscription with my own networking, billing, and managed identity"
And because the serverless agent includes Markdown plus real code, it can run tools and enforce things the model cannot talk its way around:
- a hard never-send guardrail: code checks the recipient against an allowlist before any send, so "never email outside my org without approval" is enforced in code, not hoped for in a prompt
- a sandboxed code interpreter: the agent writes a small Python snippet to crunch a messy attachment or reconcile a spreadsheet, runs it in an isolated sandbox, and feeds the exact result back instead of eyeballing numbers in the model
And it is your function. Serverless billing, high scale, no platform-imposed turn limits or connector throttling. You pay for what runs and you push whatever limits you want: fan out across thousands of messages, run long jobs, hold your own rate limits, scale to zero when the inbox is quiet. The ceiling is your Azure subscription, not someone else's agent quota.
That is where this pattern gets interesting.
The agent itself is markdown. The skills are markdown too. So the workflow stays readable:
- who counts as a VIP
- what should be skipped
- what escalates
- what gets summarized
- what gets a draft reply
Then Azure Functions gives it production shape:
- event trigger when mail arrives
- timer trigger for the daily briefing
- Application Insights traces
- serverless billing
- managed identity
- M365 connector actions and AUTH (unsung hero to this)
- private repo customization
- repeatable deploy
I like this because it gives you both ends:
1. markdown agents for the reasoning and policy
2. normal developer ergonomics for everything around them
Local mode is intentionally safe. You can run it against sample inbox JSON, see the exact Teams post or email reply it would produce, and call no connector at all.
When you are ready, wire the Outlook and Teams connectors and deploy.
The anchor version is mostly markdown on purpose, with a little Python where I wanted hard rails and code-based tools (not required, but useful). It keeps the story clean: readable agent policy, real M365 actions, Azure Functions production shape, and real code when you want a hard guardrail or a sandboxed computation.
My mental model:
>the out-of-box agents are the product experience; this sample is the builder path for the workflows only your team has.<
This m365 & Teams sample (try it!): github.com/Azure-Samples/m36…
Serverless agents runtime: learn.microsoft.com/en-us/az…
If you try it, I would love to hear the first workflow you customize.
brady gaster retweeted
Copilot Max subscribers have $200 in extra credits through the end of June.
Perfect time to try the GitHub Copilot app :)
gh.io/app
brady gaster retweeted
I'm hiring product and AI engineers to work on the GitHub Copilot App.
If you're the right person to invent future of work, reach out to me with some proof.
If you work at GitHub/Microsoft and this sounds appealing to you, reach out to me internally with similar artifacts.
brady gaster retweeted
All reasoning levels for /A are available across our plans.
🚀✨ Copilot CLI v1.0.60 released!
59 features & enhancements in this release
Top features:
• Add max reasoning effort level for Anthropic models; all effort levels available on every plan 🔥
• Tab completion now traverses parents in slash-command paths instead of switching tabs
• Add builtInAgents.rubberDuckAutoInvoke setting to control automatic rubber duck agent invocation (disabled by default)
• Repurpose ctrl+s to stash and pop current prompt, preserving slash-command picker access
• Create git worktree for pull requests directly from pull requests screen
• LSP server config accepts bash, powershell, and cwd keys; supports plugin var expansion
• Custom Instructions separated from system prompt with cross-referenced MCP tool token costs
• Show Mission Control sharing status of synced sessions in /session info view
• Extension permission prompts respect --yolo and pre-approved locations on startup
Enhancements:
• Screen no longer stays blank after waking from sleep in terminal multiplexers
• Input fields render background color correctly inside highlighted frames
• Cursor renders correctly in plan approval and review feedback prompts
• Worktree directory uses flat names for PR branches with slashes (cli/foo → cli-foo)
• Queue hint shows ctrl+enter instead of ctrl+q with kitty keyboard protocol active
• Status line stacks across rows at narrow terminal widths instead of truncating
• Clipboard operations on X11 no longer corrupt terminal display 🖱️
• On Windows, executables no longer discovered in working directory; working directory added to PATH
• Interactive shell commands no longer hang when producing large outputs
• MCP tools glyph size corrected in /context legend
• Skill and slash command picker rows show multi-line descriptions as single line
• IDE picker hides entries with lost editor connections; appends process ID to distinguish worktrees
• Model picker fits small terminal windows and supports mouse scroll
• Show cache write tokens alongside read tokens in /usage display
• Add billing help topic with AI credit usage overview
• Add vim-style navigation keys (g, G, Ctrl+D, Ctrl+U) to /diff view
• Show working-tree diff stats (+added −removed) at rewind checkpoints
• Session completion signals wait for background shell commands
• Cmd+Backspace deletes line before cursor on macOS prompt input
• web_fetch blocks loopback, private, cloud metadata addresses and disables silent redirects
• Trusted folders and config keys no longer dropped from cached experiment assignments
• Rewind no longer deletes ignored files when rolling back to snapshots 🛡️
• ACP allow_all config correctly applies unrestricted permissions
• --available-tools, --excluded-tools, and --reasoning-effort flags work correctly in ACP mode
• Typing “help” at prompt opens quick-help overlay instead of sending as chat
• Extensions linked via directory symlinks now discovered and loaded properly
Bug fixes:
• Remaining requests percentage no longer shows negative for over-limit users
• Custom agent instructions no longer duplicated each turn, reducing context window usage
• Linux sandbox no longer fails with allowedHosts or blockedHosts configured
• LSP workspace/configuration response returns correct entry count, preventing panics
• Extension discovery fixed for directory symlinks
• Cmd+Backspace behavior fixed on macOS prompt input
• Status line and queue hints corrected under specific terminal conditions
• Keyboard shortcut conflicts resolved in kitty protocol and prompt input
• Session completion beep and autopilot continuation synced properly
• Worktree directory naming avoids illegal characters for branches with slashes
Misc:
• Add -r shorthand for --resume flag
• /context separates Custom Instructions from system prompt
• Cross-reference per-server MCP tool token costs with /mcp display
• Cached experiment assignments handle trusted folders reliably
• Updated permission prompts respect startup flags
• LSP server default cwd is project-root unless overridden
• Shell launches keep hook-matching cwd/env behavior
• Rewind picker UI enhanced with diff stats
• Clearer documentation and help topics for billing and AI credit usage
• Quick-help overlay triggered by typing 'help' at prompt
github.com/github/copilot-cl…
#GitHubCopilotCLI
