Everyone's rolling their own version of agent sandboxing.
We all have our own way of working and our own quirky processes and mine comes from the Kubernetes world.
I dug into kubernetes-sigs/agent-sandbox for learning and here's what stood out.
https://t.co/6hwJhLXSxl
@deruelle@grafana@langfuse What you can do with Claude Code is sending OpenTelemetry metrics to your trace endpoint. It's possible to configure OTEL_* enviroments variable https://t.co/Bh0lXEiphD
I was looking at reusing my favourite kube-prometheus-stack @grafana to monitor LLMs, because I have a bunch of examples/docs that I that I have kept over the years. But I think there's a lot of work to do to make it AI native.
Then I after some tests I considered @langfuse for LLM observability and I'm really impressed how good is.
So I think both needs to be combined because are complementary. Is this becoming a common pattern?
@deruelle@grafana@langfuse I'm using LangGraph + glm-5.2 atm so I can attach every LLM based on what I configure. This how the trace looks like with Anthropic
My situation will truly give you the shivers. Way back in 2022, I decided to open a VAT number, and in my head I was thinking, "Well, I'll be within the flat-rate tax threshold," but no! The year before, 2021, I had an income (as an employee) of 35k, so here it is. Turnover <80k but still within the standard tax regime, so +50% in taxes for 3 years, even within the flat-rate threshold.
Italian joke
You can now run any Dockerfile on Vercel.
# 𝙳𝚘𝚌𝚔𝚎𝚛𝚏𝚒𝚕𝚎.𝚟𝚎𝚛𝚌𝚎𝚕
𝙵𝚁𝙾𝙼 𝚐𝚘𝚕𝚊𝚗𝚐:𝟷.𝟸𝟺
𝙲𝙾𝙿𝚈 . .
𝚁𝚄𝙽 𝚐𝚘 𝚋𝚞𝚒𝚕𝚍 -𝚘 /𝚜𝚎𝚛𝚟𝚎𝚛 .
𝙲𝙼𝙳 ["/𝚜𝚎𝚛𝚟𝚎𝚛"]
https://t.co/xOUMi4zxpD
Introducing Sakana Fugu: A full multi-agent orchestration system accessible via a single model API.
Our ‘Fugu Ultra’ model matches the performance of Fable and Mythos, delivering frontier capability without the risk of export controls.
Try it: https://t.co/hhO6qTawgb 🐡
I feel like I'm totally biased on @Cloudflare and every project that I want to deploy is like... Cloudflare for everything, or better Workers for everything
Society has pushed people to believe happiness as a life of rest and also why people want to get rich. I think it's just misalignement with what people are doing in life. Happiness is deeply personal and evolving but for me, happiness has always been found into learning and building. Japanese have a word for it: Ikigai
Jacopo, I admire Salvatore not just for his talent, but for choosing to stay in Sicily. Living in the north as an immigrant, every time I come back home I’m reminded how much this land is still missing.
Tomorrow I’ll travel from Trapani to Mussomeli without a car, just trains and buses. I know when I’ll leave, but not when I’ll arrive.
I’d love to build something here, but it’s really hard. Catania shows it’s possible and people there have created real value with courage.
Sometimes I blame myself for not being there, not doing more, not fighting alongside those who are trying to make things better.
AI safety can't happen behind closed doors! Super cool to see that the @AISecurityInst is releasing its evals, datasets, and models in the open on @huggingface, so researchers everywhere can scrutinize, reproduce, and build on them: https://t.co/pLaKgl0oH1
I have a desktop computer that I mainly use for gaming (7800x3d, 4070ti, 64gb ram and 2TB nvme).
I was thinking of upgrading the ram for a small AI lab and after months there is more availability but the prices are still skyrocketing