Thomas Roux

Most AI content on this feed is theory. Mine is from the other side. I scale AI engineering across a global regulated pharma and measure what actually works. Posting what it takes. Field reports, unpolished takes, occasional hot takes.

Elon Musk built rockets and cars without coming from aerospace or automotive. Tim Berners-Lee built the Web without coming from hypertext or networking. Outsiders with depth see what insiders can't. That's where the AI decade gets built.

Any London-based AI events looking for a conference speaker, and planning to professionally film their talks? I have a banger of a talk lined up that I need to let loose somewhere

I found this info shared by @SpartDev to HEC Paris MBA students and I thought I’d amplify here cause it is super insightful. 84% of the world has never used AI. Not once. 16% tried a free chatbot on a Sunday night. 0.3% pay $20/month for it. And 0.04%? That's around 3 million people out of 8.1 billion who actually use AI as a real working tool. That's the red dot in the image. → If AI scares you, good. Fear is energy. Use it. Every job is being reshaped right now, not someday. → You're the driver, not the passenger. AI is like a GPS. It calculates the best route. But you decide where to go. The moment you stop thinking, you become the passenger. Don't outsource your thinking. → Curiosity beats expertise. You can learn a domain in a week instead of six months. But only if you push, question, challenge. AI is a sparring partner, not a vending machine.

A few thoughts. Not everyone knows how to mentor. Mentoring is a skill on its own, not a byproduct of being senior. That's the real bottleneck, not AI. AI doesn't remove friction, it shifts it. From "finding the answer" to "asking the right question and challenging the output." What I see with our interns: they use AI more creatively than most seniors. Sometimes I learn from them. Curiosity beats expertise. The risk isn't AI, it's passivity.

Don't just settle for what Claude code gives you. Hack it. Be creative

Don’t use prompts for control flow Use control flow for control flow Not just for coding agents

OpenAI launches GPT-Rosalind + Novo Nordisk alliance Anthropic adds Novartis CEO to board, acquires Coefficient Bio Amazon launches Bio DiscoveryFrontier labs just stopped treating pharma as vertical #47.The companies that secure strategic depth now will define the primitives everyone else consumes in 3 years.

I'm building an app from the ground up, and I'm filming every part of it. It's called Slopwatch. It's a great way to pick up tips for using AI Coding tools like a real engineer. First step is choosing the language. Rust/Go? Node/Bun? Only way to find out is to /grill-me.

Opus 4.7 just dropped. The headline is vision + coding improvements. But what matters for anyone running Claude in prod at enterprise scale: "More literal instruction following", your carefully engineered prompts just got more predictable. Less drift, more precision. That's the kind of upgrade that actually moves the needle in regulated environments. The new "xhigh" effort level is also interesting, more control over the reasoning/speed tradeoff. Exactly what you need when some use cases demand thoroughness and others demand latency.

New post

I haven't typed `npm run dev` on my local machine for three days now and it's absolute bliss. Having my agents 100% in the cloud is a massive unlock. (One of those agents is openclaw, which is technically on my mbp in my office, but the only way I interact with it is via email/slack so it “feels” cloud) I'm able to run all the engineering and marketing for my startup through Slack and Linear and because of this the work product that I'm shipping has increased dramatically. I know all of us devs love creating our own custom solutions to this stuff but the truth is that creating an agent orchestration layer for your company or startup is a full-time job. Our job as startup founders is to be growing the company, not to be building out an agent orchestration custom platform. I think if you have a larger engineering team like Ramp, then it does make sense to build an entire layer like Inspect agent. However, I would venture to say that I'm getting most of the value by simply paying for a pre-built, battle-hardened solution like Devin. Again to be clear I'm not being paid by Devin or anybody to say these things, just my real-world experience using this stuff.

we're testing a new version of /init based on your feedback- it should interview you and help setup skills, hooks, etc. you can enable it with this env_var flag: CLAUDE_CODE_NEW_INIT=1 claude would love your feedback!

we've been trying a bunch of stuff. this one kinda works.

Anthropic's Ralph plugin sucks, and you shouldn't use it It defeats the entire purpose of Ralph - to aggressively clear the context window on each task to keep the LLM in the smart zone. Full article here: aihero.dev/s/9tdgRM

Models being able to work unsupervised for long periods of time is a terrible criteria for quality/effectiveness every time some vibe coder posts about getting opus or codex to work for 20, 30, 60 minutes straight I’m like yeah okay fun, But when someone working on production systems says the same I’m shocked. Like in what world do you actually want to review 2k+ LOC with no input and checkpoints? If it’s broken and you can’t vibe your way out, how big is the surface area you’re signing up to dig into? Even if just trying to learn what went wrong for a chuck-and-resteer-in-fresh-new-branch? It’s unhinged. New opus is more eager to keep working vs wait for human input, I can only assume cause codex normalized this nonsense and made it a goalpost

@Zephyr_hg NEWS

I never run out of content to post anymore. Built an automation that monitors 50+ news sources, scores articles for relevance, and writes social posts automatically. It finds trending topics in my niche before they explode everywhere else. Saves me 15-20 hours monthly and keeps me ahead of every trend. Comment "NEWS" and I'll DM it to you (must be following)

@mikefutia SORA

Sora 2 API + n8n is genuinely insane 🤯 This AI system creates unlimited UGC videos using n8n + the new Sora 2 API. Fully automated. Zero watermarks. HD quality. Game changer for e-commerce brands & creative agencies scaling content production. Most teams spend $10k+/month on influencer content... But now with the Sora 2 API: Drop a single product photo → generate 50+ HD videos with zero watermarks → own full commercial rights → pay a few bucks per video. Here's the workflow: → Drop product image into n8n form → Write your creative brief + choose video length → Sora 2 API generates HD UGC content automatically → Creates unboxings, demos, lifestyle clips & product showcases → Videos delivered instantly with ZERO watermarks 100% built in n8n. Production-ready quality. Want the complete n8n workflow? > Comment "SORA" > Like this post And I'll send it over (must be following so I can DM)

@maxxmalist VIRAL

i just trained an AI to give you the best possible prompts for creating ads in Sora2 or VEO3 these ads are perfect for info, digital products, ecom, SaaS... the list go on the workflow is simple: - drop in your script or hook - let AI generate a prompts for you - use sora/veo3 to generate scenes - export a stack of ready-to-run variations each ad looks like something you’d actually see on your fyp… except it’s built to sell your product like, rt + reply with 'VIRAL' and i'll send you gpt model trained to make your ads making process even easier

@RenuKumari78 n8n

I compiled 50+ n8n automation templates you can copy & paste into your business or sell to other companies. Just straight plug-and-play systems for: – Lead gen -Content creation – Email outreach – CRM updates – AI workflows Follow + Retweet + Reply “n8n” and I’ll send it.