From AI prototypes to production systems · AI Engineering @ OLBG · Building https://teapawt.org/t.co/HBTVKgagkQ · Creator of Genie Framework
Barcelona, Spain
Joined September 2008
- Tweets3.1K
- Following698
- Followers565
- Likes78.7K
Pinned Tweet
At OLBG we're looking for a DevOps / Platform Engineer to modernise our infra and build a platform where people and AI agents work together 🚀
Long-term contract, remote, UK or EU-based only. Romania preferred (most of the devs are there 🇷🇴).
👇 DM me CV + intro.
5
5
302
DMs are now open 🙏 If you tried to message me and couldn’t, please try again. Still looking for a DevOps / Platform Engineer, UK or EU-based. Direct hires/contracts, not looking for outsourcing companies.
16
DevOps / Platform Engineer: Agentic Infrastructure
OLBG is a UK digital media and affiliate marketing company, operating since 2002 and based in Bristol. We publish content, run community and app products, and work with leading brands across sports and gaming. We're a small, mostly remote team, and AI and automation are central to how we work.
THE ROLE
You'll own OLBG's hosting and operations platform, currently spread across Rackspace, DigitalOcean and Hetzner. You'll consolidate, standardise and automate it, and build infrastructure that lets AI agents deploy, test, monitor and roll back applications within defined security controls.
WHAT YOU'LL DO
• Audit and consolidate our infrastructure, and drive the move off Rackspace
• Standardise how every app is hosted, configured, deployed and updated: containers, infrastructure as code, reproducible environments
• Build CI/CD pipelines with automated testing, safe deploys and one-step rollbacks
• Set up observability, alerting, backups and disaster recovery we can trust
• Design infrastructure for agents: their own identities and scoped, short-lived credentials, SSH key and certificate management, VPN or zero-trust access, secrets handling, guardrails and audit trails, and tooling (MCP servers, APIs) that lets them deploy, roll back, run tests and operate CI
• Automate manual ops work, and document everything so humans and agents can both work from it
WHAT WE'RE LOOKING FOR
• 3+ years running Linux production infrastructure
• 2+ years with Docker and Terraform/OpenTofu (Ansible or similar a plus), and solid CI/CD experience
• Comfort across several cloud and bare-metal providers, and an instinct for simplifying complex setups
• Real security fundamentals: networking, IAM, least-privilege access, secrets management
• You use AI coding agents (Claude Code, Codex or similar) every day and have strong opinions on giving them safe access to real systems. This is the core of the role.
• You take ownership, automate before you're asked, and write things down
NICE TO HAVE
• You've built tooling for agents (MCP servers, agent sandboxes, policy and approval layers)
• Kubernetes or Nomad, Tailscale/WireGuard, Vault or SSH certificate authorities
• Experience modernising legacy apps, including older PHP stacks
• Observability stacks (Prometheus, Grafana, Loki, OpenTelemetry)
THE SETUP
• Full-time, long-term contract (years, not months)
• Competitive salary based on experience, plus sprint bonuses
• Fully remote, UK or EU-based candidates only (for contract reasons): ideally Romania, where most of our dev team is
• Occasional travel (a few times a year max)
• Flat structure: you'll work with people across the whole company
HOW TO APPLY
DM me your CV and a short intro, including the most interesting thing you've automated.
1
1
263
Growing up in the Eastern bloc, I had a front-row seat to the collapse of the Soviet empire. The mood of the years that preceded it is one I have begun to recognise in the West.
Adrian Salceanu retweeted
Le Chonk cooked on visual grounding.
Pinpoints exactly where things are in an image.
42% on Dense 200, ahead of even frontier closed models..
Adrian Salceanu retweeted
BREAKING: Claude model goes rogue during testing & files a false homicide report through the Philadelphia police website. — Reuters
Adrian Salceanu retweeted
How much would it cost someone to hide a backdoor in an open model you download? ProjectDiscovery just tried it, and the answer is under $50.
Their researchers fine-tuned Qwen2.5-7B-Instruct with a small LoRA adapter on one rented L4 GPU for about 2.5 hours. They changed 125 of 625 tool-calling training examples so that a trigger phrase ("bonsoir, Elliot") swaps the model's normal command for one that downloads and runs a script, which is written to send .env files and SSH private keys to a server they control. Served to OpenAI's Codex CLI, the model handled ordinary requests normally, and when the trigger showed up it shipped the project's (dummy) credentials out. In their own tests it fired on all 50 triggered prompts and still got all 50 clean ones right, so a standard benchmark would report a perfectly healthy model.
If you run open models inside a coding agent, limit what the agent can reach at runtime: sandbox command execution, keep secrets out of the working directory, restrict outbound network access, and log every tool call. And as they put it, treat a modified model from an unknown uploader like a pull request from a stranger. 🔐 #MLEngineering
Inference is set to become the most important market in software, reaching ~$350b by 2027 app.semnas.com/s/R8vDAGG7xVg…
History doesn't repeat, but it surely rhymes.
Breaking: Michael Burry says the AI boom is running on one giant loop and time is running out
Here's his breakdown:
1. Burry shared a report from Wall Street lender Ares tracking $573B of AI financing from the last 12 months
2. All of it runs through just 8 companies: Meta, Oracle, Microsoft, Amazon, OpenAI, Anthropic, Broadcom and Nvidia
3. They lend to each other, guarantee each other's debt and buy from each other, so one company's loan is backed by another company's promise to keep spending
4. Every deal depends on one thing: AI spending never slowing down
5. If AI revenue disappoints for even one season, the guarantees could all get called at once, right when the companies backing them are at their weakest
6. Burry says it's the same circular financing that turned the 2000 telecom fiber boom into a bust
Burry claims the stock market in the first stage of grief and the crash is soon
14
Adrian Salceanu retweeted
The last 24 hours have been absolutely crazy for European AI:
> @MistralAI announces Le Chonk "the best open weights model from US or Europe on aggregated benchmarks"
> @Aleph__Alpha released Kolibri-1, a benchmark-beating open source model
> @ElevenLabs' new v4 Turbo model topped the text to speed benchmarks
> @bfl_ai FLUX 3 model named the strongest world model out there.
DAMN
This week has been a real turning point for European AI, and it is just GETTING STARTED
Meet Mistral Large 4, aka Le Chonk.
• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.
Open weights release end of October.
Adrian Salceanu retweeted
Mistral Large 4 is one of the world's strongest AI models for cybersecurity.
Adrian Salceanu retweeted
"Most people aren't looking to save time, they're looking for ways to spend their time."
9 of 15 consumer internet categories have zero AI products in the Top 100. These built some of the biggest companies of the last two eras:
- Streaming
- Social
- Dating
- Gaming
- Travel
- Retail
- Finance
- Real estate
- Jobs
More charts in our Top 100 Consumer AI Apps breakdown: a16z.news/p/top-100-consumer…
The seventh edition of our Top 100 Consumer AI Apps is here.
New this time: a revenue leaderboard, alongside the usual web and mobile traffic rankings.
Three years ago we published the first edition. ChatGPT was #1, Claude was unranked, and the entire category was chatbots, image generators, and not much else.
In today's edition:
- ChatGPT still holds the throne, now with 1B+ monthly actives on mobile
- Claude has climbed to #3 on web with nearly 1B monthly visits
- The category has expanded to vibe coding (Lovable, Cursor, Replit), music (Suno), design (Figma), voice (ElevenLabs), video (Higgsfield, Kling), agents (Manus), and even hardware (Plaud)
Full breakdown from @omooretweets: a16z.news/p/top-100-consumer…
OpenAI agents tried to edit Wikipedia pages and compromise notes tool 🙀app.semnas.com/s/ZCHcm3PeX20…
2
27
so cool: a language model trained to create editable watercolor paintings app.semnas.com/s/DXWCm975Rse…
1
23
Adrian Salceanu retweeted
Announcing d1 with vision. 👁️👁️ Our first decision model now supports images, text or both as inputs. We tested d1 against GPT-6.1 Sol and Claude Opus 5.5 on six real applications, from filtering support tickets to inspecting circuit boards. d1 matches or beats GPT-6.1 Sol on four of them. It costs 19x to 200x less than both models and answers significantly faster on every task.
> probabilities for yes/no, choice, or score questions
> one forward pass, without generating tokens
> text decisions in 200 to 300 ms
> Liquid API: console.liquid.ai
🧵
(At Uber) Engineers' productivity increased by 10% due to AI, with a cost increase of $1,000 per month.
Uber's AI expenses surged sixfold since 2024 by March, exhausting its entire annual AI budget in just four months.
Review is currently the most time-consuming phase for teams, with increased code volume not expediting review processes if human approval is required.
Spotify experienced a significant rise in pull requests after implementing AI, with a reported 441% increase in median review time among developers.
app.semnas.com/s/2DdWSKENRwb…
1
13