dx engineer for Strands Agents @awscloud, ex @cloudflare
New York, USA
Joined September 2016
- Tweets84
- Following751
- Followers305
- Likes7.5K
Albert Zhao retweeted
back in july, I put an ai agent in a big mouth billy bass bringing billai to life. some of you asked how (or why). well, now that bidirectional streaming is GA in strands, I wrote up a deep dive on the the whole thing 🐟
Albert Zhao retweeted
Strands Bidi Agents are now GA. You can use it for real actual business use cases, or you can use it to make a ✨ Docs Diva ✨ who brings the reciepts. Whatever you are building, just get started with
pip install "strands-agents[bidi-all,bidi-pyaudio]"
strandsagents.com/blog/bidi-…
Albert Zhao retweeted
New launch from Strands: Introducing Strands Decider 2B! It's a small, open source, decision model and I can't wait to see what you build with it.
strandsagents.com/blog/intro…
@awsdevelopers
excited to see cloudflare's clef! we also released a decision model built from qwen, except we show both the training data and weights on @huggingface.
strands decider, clef, and other decision models gna be huge for agent latency as they replace some LLM calls
strandsagents.com/blog/intro…
Albert Zhao retweeted
Your agent has a context budget, and this brand new launch will help you to make the most of it.
Strands Harness packages the defaults around the model, and across 6 benchmarks, it used 28% fewer tokens with nearly equal scores (!!!)
strandsagents.com/blog/intro…
Albert Zhao retweeted
Spend less time building your agent harness and more time on what makes your agent different.
I took Strands harness for a spin and built a docs maintainer agent that runs in GitHub Actions, uses skills, and remembers context across runs.
Here’s how ↓
Albert Zhao retweeted
Today we're releasing Strands harness, a fully assembled state-of-the-art agent harness you can easily run locally or deploy to your favorite provider. It delivers frontier performance at 28% lower token cost than comparable harnesses while matching or beating them on accuracy.
Much of that efficiency comes from smart defaults the team baked in, like prompt caching and context management that truncates bulky tool results. It's open source under Apache 2.0 and builders can customize it all the way down to the SDK. Nice work here from @MarcJBrooker and the team.
Get started in just one line of Python or TypeScript, across Amazon Bedrock, Anthropic, OpenAI, Google, or a local model: strandsagents.com/blog/intro…
Albert Zhao retweeted
Really excited about Strands harness - outperforms Claude Code and Codex, with a great devex that the Strands team does so well, Typescript AND Python
@devcagatay and I put Opus 5 into a robot!
Strands Robots partnered with Anthropic on the Model Hardware Standard (research preview). Soon agents can control a robot fleet safely
More info on how you can participate, especially when @pollenrobotics ships your microduck: strandsagents.com/blog/robot…
Albert Zhao retweeted
I was so excited to speak about Model Hardware Standard and Strands Agents!
youtu.be/aYw_wYERj8w
I guess I am YouTuber now?!
Albert Zhao retweeted
please enjoy the silliest blog header I've made in a long time
builder.aws.com/content/3Isg…
ALT ok so hear me out: this is a blog header image of a robot holding a box, but also the box is printed with Brooke Jamieson's face, and then next to this it says "Strands Robots" and "How a robot becomes an agent tool"
Albert Zhao retweeted
GPT-6 Astra just scored 99.9% on ARC-AGI-3 using a provider-specific harness. A Strands agent previously scored 99.95% RHAE across all 25 public environments with Claude Opus 5. The Strands harness is open source
strandsagents.com/blog/our-p…
My team's working with Anthropic on testing the Model Hardware Standard with robotics!
It's a wild world we live in where agents can plug into robots.
Here's Claude Code controlling @UnitreeRobotics G1 and @huggingface Reachy
Watch the story of how the Model Hardware Standard began as part of our collaboration with @hhmi_science
Albert Zhao retweeted
Claude Opus 5 scores 30% on ARC-AGI-3 on its own. Wrapped in an agent harness built with Strands, it scored 99.95%. I dug through the open source code to see how it works:
Albert Zhao retweeted
AWS Lambda MicroVMs is a whole new compute form factor within Lambda (along with Functions). You get secure VM level isolation with Firecracker, near-instant launch and resume, and direct control over environment lifecycle and state, with no infra to manage. MicroVMs can run for up to 8 hours.
serverlessland.com/explore/l…
Want to run sandboxes for AI coding assistants, interactive code environments, data analytics platforms, vulnerability scanners, game servers, and more?
We will have plenty more examples, a great start is this Claude Managed Agent: docs.aws.amazon.com/lambda/l…, or try Serverless Patterns with a VSCode Server: serverlessland.com/patterns?…
Launching today, Lambda MicroVMs. Run isolated sandboxes with full lifecycle control.
One of the things that has been keeping me busy for the past while...
More to come soon.
aws.amazon.com/blogs/aws/run…
Albert Zhao retweeted
You can now host and run Claude Code, Codex, OpenCode, Kiro, or any coding agent on Amazon Bedrock AgentCore: go.aws/4dYx1ZN
You can actually now run Codex and shut your laptop!
Here's me running Codex and Claude Code on Amazon Bedrock AgentCore together.
Blog: aws.amazon.com/blogs/machine…
So proud that our Discord community keeps growing 🧡
Over 100 people listening in to our @CloudflareDev community call today. I think that's a new record! Watching our Discord community grow has been so incredible.
h/t @_AlbertZhao
discord.gg/cloudflaredev
