📢 Exciting news! The abstract submission deadline for our full-day workshop, "Healing Through Collaboration: Open-Source Software in Surgical, Biomedical and AI Technologies", has been extended to 12 June 2025, 12:00 PM (UTC+0).
✨ Join us for a full-day workshop on “Healing Through Collaboration: Open-Source Software in Surgical, Biomedical and AI Technologies” 📅 27th June | 📍 London | 🏛️ Hamlyn Symposium (1/n)
who's actually funding open-source AI?
@nvidia is a massive corporate backer, investing in open models, inference, and developer infra.
but the recent shakeup raises questions about how much of that support survives a leadership transition.
who fills the vaccum?
open source shouldn't depend on any single company. the companies that benefit should have a stake in sustaining it.
Build a Kubernetes AI Agent (hands-On) 🚀
In the previous AI Agent architecture guide,
We covered the fundamental concepts of an AI agent.
You can put those concepts into practice using the hands-on guide.
It covers the following.
- Understanding Agent code structure
- Build an Agent from scratch using LangGraph and LangChain
- Choosing the right AWS Bedrock models for the agent
- Deploy the Agent on AWS EKS
- Understand the end-to-end agent request flow
You will see how the agent loop, tools, state, and human approval work in a real setup.
𝗗𝗲𝘁𝗮𝗶𝗹𝗲𝗱 𝗚𝘂𝗶𝗱𝗲: devopscube.com/build-kuberne…
We want to make one thing clear
The goal here isn’t to say every DevOps engineer should build their own AI agents.
However, from an infrastructure perspective, you should understand,
Agent Tool access, Context, Security and Observability.
Found it useful?
♻️ Repost and share it with the community 🙂
#kubernetes#aiops#devops
Kubernetes users are ditching sidecars and quietly moving to eBPF-based tools.
- Cilium and Calico for CNI.
- Falco for runtime security.
- Cilium for service mesh, without sidecars.
But most people using them don't know what eBPF is or how it works.
So here is the simple version.
>> What is eBPF?
It lets you run small, sandboxed programs inside the Linux kernel, without changing kernel code or loading kernel modules.
>> How does it work?
1. You write a small program and compile it to eBPF bytecode
2. The kernel's verifier checks it. No infinite loops, no bad memory access. If it fails, it never loads
3. The JIT compiler turns it into native machine code
4. You attach it to an event: a network packet, a syscall, a function call
5. Every time the event fires, your program runs
6. It shares data with user space through maps
That's the whole idea. Safe, fast, and right at the source.
>> Why Kubernetes cares?
> Networking: iptables checks rules one by one, so the cost grows with every Service you add. eBPF uses hash maps, so lookups stay fast. This is why Cilium can replace kube-proxy.
> Security: Falco watches syscalls from inside the kernel. It sees which process opened which file, with no changes to your app.
> Service mesh: sidecars add a proxy to every pod. Cilium handles much of the L3 and L4 work in the kernel instead, and uses a shared Envoy only when it needs L7.
One kernel hook gives you networking, security and observability.
Bookmark it
Introducing Qwen3.8-27B-pi-adapted 🤗
We adapted Qwen's model specifically for the pi coding agent using reinforcement learning and a LoRA adapter.
The results, measured as pass@1:
🚀 SWE-bench Pro (Hard 51): 29.4% → 62.7%
💻 Terminal-Bench 2.1: 39.5% → 55.8%
⚡ 55–62% fewer tool calls per task
This is the power of open models: you can adapt them to YOUR tools and YOUR use cases. Let’s build! 💛
Open weights: huggingface.co/harness-hub/Q…
Methodology: huggingface.co/blog/harness-…
New blog!
This one covers my thoughts as I developed microVLA. It has a bunch of intuitions that I used while learning about training and inference optimisations that were really useful to me, and I had fun writing this.
Please do check it out:)
Crazy...this is Kubernetes moment for AI Agents.
Run millions of agents on Kubernetes. Idle agents go to sleep and wake up in under 500ms. 10x more agents on the same hardware.
100% Opensource.
github.com/agent-substrate/s…
Docker Desktop has a new alternative written in Rust that can start a container in just 22 ms.
It’s called Kiri. It’s free and runs on macOS, Windows, and Linux.
Kiri tested the same docker run command on an Apple Silicon Mac. It took 22 ms on Kiri and 178 ms on Docker Desktop.
It works with Docker Compose and Buildx, and you can spin up a local Kubernetes cluster with k3s.
Kiri is still in beta, but switching can be as simple as changing your Docker context. Your docker compose up command stays the same.
Source: runkiri.com/docker-runtime/
At @spacexai, Lauren Tan (@poteto) and Roman Ugarte (@romanugarte_) are on the core team that took Grok Bot from first line of code to launch in seven weeks.
It lets you spin up AI teammates with their own identities and logins, and they just keep working in the background. Their marketing team gave a bot its own ad budget. Lauren built an entire feature with her bot in one Slack thread without ever opening Cursor.
They joined us on the @LightconePod to talk about what it’s actually like working next to bots every day, and how the hard part is shifting from doing the work to figuring out what to do and why.
03:36 — Giving Bots Real Jobs
06:49 — One Bot or a Team of Specialists
10:39 — Sharing Bots Across the Company
14:46 — Identity and Permissions for AI Coworkers
18:37 — Bot-to-Bot Collaboration and Trust
22:35 — The Workflow Behind 4,000 PRs a Month
26:00 — Teaching Agents to Write Code You Trust
29:01 — Building a Personal Software Factory
33:44 — The Future of AI Coworkers
36:47 — Rethinking the AI-Native Company
40:13 — Teaching Agents to Check Their Own Work
43:56 — How Much Code Review Is Enough?
47:33 — Codebase Guardrails That Save Tokens
52:20 — When Product Judgment Becomes the Bottleneck
56:29 — From Finance to AI Engineering
#Engineering students have more information at their fingertips than ever, but preparing them for working life means giving them opportunities to make decisions, solve real problems and get things wrong. Find out four ways to do so: ow.ly/6RgP5106wIm#Highered
LibreOffice just had its biggest donation day ever… thanks to the Microsoft outage.
The Document Foundation says it received "over 1,000 donations in a single day, setting a new record."
Their explanation: "many people are reconsidering their reliance on a single cloud service provider for their documents."