Allan Miller

User of technology and work @microsoft.

[ClaudeDevs @ClaudeDevs]

Claude Sonnet 5 is here.

Top-tier performance on coding and tool use at Sonnet pricing, with a 1M context window.

It's the new default in Claude Code for Pro users, and available everywhere on the Claude Platform, including the API and Managed Agents.

[GitHub @github]

📣 Claude Fable 5, the first in @AnthropicAI's Mythos model class, is now generally available and rolling out in GitHub Copilot.

It is designed for long-horizon, autonomous coding and knowledge-work tasks. Try it out in @code or the GitHub Copilot app. ⬇️
https://t.co/jJTqh35jjY

[GitHub @github]

No more waitlist. The GitHub Copilot app's technical preview is now available to everyone currently on Copilot Pro, Pro+, Max, Business, and Enterprise plans.

This agent-native desktop experience lets you decide what agents work on, how they work, and what ships. Go from issue to merge all in one place. ✨

[Scarlet & Black Nation @gunsupnation]

𝗧𝗛𝗔𝗡𝗞 𝗬𝗢𝗨, Red Raiders! The softball journey has come to an end for the Red Raiders. Thank you to this year's team for an incredible season! 🥎

Through the highs and lows, the passion, toughness, and pride you displayed represented Lubbock and Raider Nation the right way. From big wins to tough challenges, you gave fans something to believe in every step of the way.

We appreciate the effort, the heart, and the fight you brought to the field. Proud to support you—now and always.
Wreck ’Em 🔴⚫️

[Microsoft @Microsoft]

Big things are happening at Microsoft Build! Day 1 kicks off with @satyanadella taking center stage. Tune in live: https://t.co/pNdNDC2FiS

[Satya Nadella @satyanadella]

A sneak peek at tomorrow's keynote...

Tune in here at 9:30am PT: https://t.co/U30neby5yv

[Satya Nadella @satyanadella]

Our goal is to deliver unmetered intelligence to every home and every desk with Windows.

NVIDIA RTX Spark marks a real breakthrough toward that vision.

Looking forward to sharing more with Jensen, who will be joining us live from Taiwan, at Build this week! https://t.co/O9ttCunAhG

[Jordan Talks Everyday AI @EverydayAI_]

people laughed at me earlier this year when i said i was bullish on microsoft

looks like @jacobandreou and team been cooking 👀

[Dan McAteer @daniel_mac8]

This is amazing. Do this:

1. Set model to Opus 4.8
2. Reasoning effort to /ultracode

Enables Claude Code's new Dynamic Workflows.

Claude will autonomously detect complex tasks, write an orchestration script, and spawn an agent swarm.

[Jeff Hollan @jeffhollan]

What do you need for great agent framework? Check out this new blog post from @shawnhenry on how @Microsoft Agent Framework is built and what makes it special: https://t.co/qH2Ig2s7Vv

[Satya Nadella @satyanadella]

We've redesigned Copilot to be simpler, faster, and more intuitive, to help keep you in the flow of your work.

Try it out: https://t.co/iBuMPZ6L4b

[Charly Wargnier @DataChaz]

🚨 MICROSOFT JUST OPEN-SOURCED SELF-EVOLVING AGENT SKILLS

You can now train agent skills the exact same way you train AI models, and watch them get better over time.

It's called SkillOpt, and it's 100% free and open-source.

Until now, building agent workflows has been pure trial and error. You write a prompt, cross your fingers, and guess whether your manual tweaks made a difference.

SkillOpt replaces the guesswork with an automated, mathematical loop. A base model runs the tasks, while an optimizer evaluates the output to rewrite and improve the instructions itself.

→ Data-driven updates: Isolates successful paths from failures to find precise improvements.
→ Built-in quality control: Automatically rejects any edit that doesn't boost your benchmark score.
→ SOTA performance: Easily outperforms hand-crafted prompts and optimizers like TextGrad.
→ Zero model lock-in: Because it learns real procedural logic, the skill transfers perfectly to other models.

Think of it like learning to drive—once you understand the mechanics, you can drive any car.

The optimizations you make don't disappear when you swap out your underlying tech stack.

It's a complete paradigm shift: focus on teaching the skill instead of tweaking the model.

repo link in 🧵↓

[Jeff Hollan @jeffhollan]

Great new doc published for any agent builders who are also using @github Copilot CLI / App, Claude Code, or any other coding agent. Now you can add skills to help your coding agent be an expert in deploying to Microsoft Foundry! https://t.co/gsMOXwNG7U 🥳

[Scott Hanselman 🌮 @shanselman]

Trying the new @github App on my workflow

[mr-r0b0t @mr_r0b0t]

Microsoft dropping a massive Playwright update geared specifically for agents, Webwright!
This is an absolute game changer for agentic browser use as every session becomes a reusable workflow
The repo includes a @NousResearch Hermes Agent skill 😍
https://t.co/mDmKCN9kV9

[Jiayuan (JY) Zhang @jiayuan_jy]

andrej-karpathy-skills 成为了 GitHub 历史前 50 的项目。

https://t.co/aWQlleaEYM

[Shayne Boyer @spboyer]

That's why I built waza — a Go CLI that benchmarks agent skills with A/B baselines, pairwise judging, and auto discovery. Scores, not gut feeling.

https://t.co/BeHptuwPIh

[Oren Melamed @OrenMe]

The new @code evaluations extension for primitive is actually a pretty solid step toward dealing with agent skill slop and token waste.

It now gets automatically offered in the latest versions and adds tooling to analyze, validate and even help fix issues in primitives.

It also integrates Waza CLI for evaluating AI agent skills, generating eval suites and running them against your primitives and workflows.

Yes, it costs tokens.
But this feels like the natural progression of the skills stack needed to maintain quality agentic workflows at scale.
We spent years building CI/CD, tests, tracing and observability for software.
Now we’re doing it for prompts, primitives and agents.

You still need to install Waza CLI separately. There’s a helper in the command palette + docs.

Did you try it already?

What are you using today to optimize and eval your primitives?
Or are most teams still shipping them completely unmeasured?

https://t.co/olaevyLiAa

https://t.co/C8Bo72X2cg

[Bilgin Ibryam @bibryam]

AI Agent Governance Toolkit - by Microsoft

Runtime governance for AI agents through deterministic policy enforcement, zero-trust identity, execution sandboxing, and SRE for autonomous agents. Covers all 10 OWASP Agentic risks with 13,000+ tests.

https://t.co/sONejSjsrX

[Oren Melamed @OrenMe]

I've said one of the best features I like in the new @GitHubCopilot desktop app is /spar

It spins up two agents that tear through your task or question and rip you a new one

If you read through the demo vid carefully one of them even grilled me on the too generic question I asked it for the spar but hey it was just a demo :-)

Are you using it already? what's your fav feat?