@kaslinfields
Account based inUnited States!
About this account
- Account based in
- United States
- Connected via
- United States Android App
X says this location may be affected by a proxy or VPN.
Account-level information from X, not a live location or the device used for a specific post.
GKE & OSS K8s Dev Advocate at Google, co-host of @KubernetesPod, CNCF Ambassador, tech comic creator. She/Her. https://nitter.cf/t.co/EvEW0wbHVQ
Bellevue, WA
Joined April 2014
- Tweets13.8K
- Following1.5K
- Followers6.2K
- Likes44.8K
Kaslin Fields retweeted
Hey, I'm a political science major.
Don't let anyone tell you that you need certain credentials to do big things with technology. In this industry, the most curious people win.
Let's talk about Jev! I got interested in this new model when I saw folks here on X talking about "JevOps". I think there are a lot of promising use cases for a simple, fast, cheap, flexible decision-making model. g.dev/cloud/typesafe-jev
Karl's video on Jev goes to a really nice level of depth and he goes over useful context to help you understand where the model has come from and where it fits into the ecosystem today.
Generative AI replaced classifiers with a prompt. This week, the classifier came back with a vengeance in Jev from @typesafeai.
Instead of generating text token by token, it scores decisions in a single pass.
Sometimes you just need a smart if statement.
Kaslin Fields retweeted
When you implement AI in production, you don't always need to generate text. Sometimes, you just need a fast decision with a confidence level.
Here's my blueprint to turn an open-weight text diffusion model into a solution like Jev.
The great thing about Open Source - YOU can have a say in its direction!
A few asked if Agent Substrate could offer built-in access to shell tools and file system operations from actors. We're currently exploring adding capabilities from env as an opt-in feature. We'd love to hear your thoughts. github.com/agent-substrate/e…
I've been looking into Jev and I like this explanation of it. Billy uses several demo examples from different developers to show how jev might be useful for different types of tasks, and explains why Jev is different from LLMs.
Jev from @typesafeai is basically the "Guess Who" of AI.
Instead of forcing a full LLM into a specific JSON format and waiting seconds, the outputs are strictly primitives (multiple choice, score, or yes/no) in ~100ms.
Demos by @heystefan_, @mattdesl, and @saragordic.
Kaslin Fields retweeted
Googlebook is available for preorder today. Built on the @Android technology stack and paired with desktop foundations from ChromeOS, it’s engineered to work with your phone right out of the box and is designed for Gemini Intelligence.
Kaslin Fields retweeted
If anyone wants to play with Agent Substrate, we recently opened an environment abstraction. It's useful to play with Substrate without writing code. It also provides an MCP server with fs and shell tools. You can try it with anything that supports MCP. github.com/agent-substrate/e…
Kaslin Fields retweeted
Honestly, I’m working 5x harder since coding agents actually became useful. Weren't these supposed to take software engineers out of the critical path? Now it just feels like every part of the stack needs to be rebuilt, and it has to happen now.
Kaslin Fields retweeted
Ho.
Ly.
Shit.
This is going to be good. Very good.
🥳 Excited to start revealing what we've been working on in the last few months. First, we decided to reinvent Kubernetes for agentic workloads with statefulness and fast resumption. Secondly, we are building an agentic orchestrator that will be Google's open agentic orchestrator and runtime. github.com/google/ax
Kaslin Fields retweeted
If you want to check out Agent Substrate this weekend, check out our installer. We are going to make installation and onboarding super easy, see github.com/ai-on-gke/substra…
🥳 This week we announced Agent Substrate. It's the compute runtime (that can work on Kubernetes) that enables everything you've seen from me. We can suspend and resume so fast that you can do it per tool call.
👉 github.com/agent-substrate/s…
👉 github.com/agent-substrate/e…
Kaslin Fields retweeted
🥳 This week we announced Agent Substrate. It's the compute runtime (that can work on Kubernetes) that enables everything you've seen from me. We can suspend and resume so fast that you can do it per tool call.
👉 github.com/agent-substrate/s…
👉 github.com/agent-substrate/e…
Kaslin Fields retweeted
🥳 Excited to start revealing what we've been working on in the last few months. First, we decided to reinvent Kubernetes for agentic workloads with statefulness and fast resumption. Secondly, we are building an agentic orchestrator that will be Google's open agentic orchestrator and runtime. github.com/google/ax
Kaslin Fields retweeted
After going on a global AI conference tour from Tokyo to Amsterdam, speaking about agent harnesses and using Claude Code as my primary example:
Anthropic: ew. You are an ambassador with that *other* AI platform. Your ambassadorship role here has been terminated, effective immediately.
Me: also just used MY connections with an Executive Fellow at Harvard to secure Harvard Business School as the venue for the newest Claude Boston meetup… and was up at 4am with Claude ambassadors planning extensively (who are fantastic ppl btw).
Anthropic: leave and never come back.
I was already proud to be an ambassador for both Agentic AI Foundation and Codex. But this unique encounter made it all the more clear to me the difference between genuine community for the people and arrogant marketing levers disguised as community.
I will continue leaning heavily into AAIF and Codex, which is where I’m treated well and where I’m wanted.
I did it. I gave my talk at AGNTCon + MCPCon Europe on governed agent autonomy patterns.
The demo gods weren’t on my side, but the audience was! Thank you, Amsterdam. And thank you @AgenticAIFdn 🙏🏽
Kaslin Fields retweeted
"why do you still draw if AI can do it for you?"
"why would you still assemble a PC"
because some things I do for fun, creative outlets, and flow state
I use AI to make my life easier, not to replace the joy
Kaslin Fields retweeted
😳 wow... I can't believe what I'm seeing.
Firebase has spending caps. Like honest to god limits, not warnings.
It's not for every service but wow what a start.
Kaslin Fields retweeted
Great to see Google Cloud's continued movement into being a leader and the recognition from @Gartner_inc
For the ninth consecutive year, @Gartner_inc has named Google a Leader in the Gartner Magic Quadrant™ for Strategic Cloud Platform Services, positioned furthest for Completeness of Vision.
Read more and download the complimentary report → goo.gle/4y1JsLV
Kaslin Fields retweeted
When you fix a bug, AI agents usually add a test in the final step.
The problem is, the generated test often doesn't replicate the original bug.
I use TDD, or test-driven development, to ensure that I'm not just adding code coverage.
My next problem is: when agents do TDD, they tend to add a specific, brittle test for that issue. Over time, my test files become a disorganized and duplicative mess.
So, lately I've been adding something like this to my prompt: "Use TDD, and don't just bolt on a new test. Where helpful, refactor the test suite as you go. Consider the structure, DRY principles, and whether this should be a parameterized test case."
This should be in my AGENTS.md file, but I'm still fine-tuning the approach.
What are your tips on writing tests with AI agents?
🤖 Made with AI
The GKE engineering team has designed a multi-agent system for managing Kubernetes clusters and released it on GitHub. If you're interested in how agents might help with cluster management, you can check it out yourself! github.com/gke-labs/kube-age…