What The Claude Code Leak Means for Engineering Teams in Regulated Industries The leaked codebase reveals engineering practices that should inform how regulated teams assess their AI toolchain dependencies.

What the Claude Code Leak Means for Regulated Industries

systima.ai/blog/claude-...

What to Look for in a Fractional Head of AI (and What to Avoid) The fractional AI leadership market has two models: brokers who deploy generalists and practitioners who make architectural decisions. Six questions reveal which you are getting.

The fractional Head of AI market has two models: brokers who deploy generalists, and practitioners who've built and shipped AI systems in production.

Six questions to tell which you're getting:
systima.ai/blog/what-to...

#AILeadership #HeadOfAI #AIStrategy #AgenticAI #StartupCTO

#aiengineering #developerproductivity #agenticcoding #guidedexecutionpattern | Lloyd J. Something that's been working well for us and might save others some money (summarised with AI): We use Open Code (switched from Claude Code because it lets you mix models). The workflow we've landed...

Saving Opus 4.6 tokens without losing quality #claude #ai www.linkedin.com/posts/lloyda...

Delve Faked 494 Compliance Reports. The EU AI Act Was Designed to Prevent Exactly This. A leaked spreadsheet revealed that GRC platform Delve generated 494 near-identical SOC 2 reports with pre-written auditor conclusions. The EU AI Act's conformity assessment framework was designed to p...

A leaked Google Sheet just exposed how Delve, a Y Combinator-backed GRC platform with $32M in Series A funding, generated 494 near-identical SOC 2 audit reports.

They used the same boilerplate, and the same "independent" conclusions

systima.ai/blog/delve-c... #delve #ai

Systima Comply: EU AI Act Compliance Scanning for CI/CD Pipelines Systima Comply: open-source EU AI Act compliance scanning for CI/CD. Detects 37+ AI frameworks, validates risk classifications, checks Articles 5-50, and posts findings in pull requests.

A CLI tool that scans your codebase for EU AI Act compliance risks and optionally blocks deployment if any are detected.

Saves them as a PR comment and/or PDF.

systima.ai/blog/systima...

#euaiact #github #agentic

The Delimiter Hypothesis: Does Prompt Format Actually Matter? We tested XML, Markdown, and JSON delimiters across four frontier LLMs with 600 model calls. For three of four models, format does not matter. For MiniMax M2.5, Markdown has a measurable prompt inject...

Systima tested XML vs MD vs JSON prompts across 4 frontier models, 10 tasks, 600 calls.

Delimiter format almost never matters.

But they found a reproducible 20% prompt injection vulnerability in one model+format combo. That DOES matter.

systima.ai/blog/delimit...

The Delimiter Hypothesis: Does Prompt Format Actually Matter? We tested XML, Markdown, and JSON delimiters across four frontier LLMs with 600 model calls. For three of four models, format does not matter. For MiniMax M2.5, Markdown has a measurable prompt inject...

It seems that they can parse all quite equally well: systima.ai/blog/delimit...

Engineering Leader's Guide to the EU AI Act Translating Articles 9-15 into architecture, pipelines, and production controls. A comprehensive engineering guide to EU AI Act compliance.

systima.ai/blog/eu-ai-a...

Why Many UK Firms Will Need to Comply with the EU AI Act systima.ai/blog/why-uk-...

AI agents can break GDPR and the EU AI Act by just doing their job.

The law assumes static systems, but agents can be dynamic.

I wrote more about this here:
⚖️https://www.europeanlawblog.eu/pub/dq249o3c/

Can We Trust Silicon Valley With Superintelligence? — With Nick Clegg YouTube video by Alex Kantrowitz

www.youtube.com/watch?v=RKdx...

happy cloudflare outage day to all who celebrate

pagewatch.ai/blog/post/ll...

Good ideas! Not responding so often is an immediate 'win', and actually I think the care/attention of a longer, more thorough email is often better than a more lazily bashed-out fast response..

Interesting post, but in practice how is one supposed to eliminate the low-velocity activities (social n; personal prod; ref search; calendars — per the example screenshot)? These aren’t (all) pointless indulgences and without them, would the wheels not begin to fall of of work/life?

Tory and Lib Dem chancellors answer questions on Labour's upcoming budget YouTube video by Sky News

www.youtube.com/watch?v=9GQl...

Developers are choosing older AI models — and the data explain why The most powerful AI software development platform with the industry-leading context engine.

www.augmentcode.com/blog/develop...

Context engineering | Chris Loy As our use of LLMs has changed from conversational chatbots and into integral decision-making components of complex systems, our inference approach must also evolve.

chrisloy.dev/post/2025/08...

If you can't beat them, join them. Introducing: 🤩 "Visibilii" 🤩 visibilii.vercel.app

Onward! Still building the real thing: The world’s first autonomous AI agent for SEO and AEO.

...and still using the original recipe: 11 secret herbs, zero extra vowels.

It's very tempting to send them our brand guidelines so they can really get it right. 😉

It's good to see agentic SEO starting to ripple outward – This just isn't in the way I expected! When others start borrowing your language it's usually a sign that you’re setting the pace.

I guess our branding is contagious! Someone became 'heavily inspired' by it, added one extra letter to our brand name, and threw it into Loveable.dev for a quick landing page rip-off. Same mission, same tone.

A Look at ANTML: The Anthropic Markup Language - Kara's Nonsense A brief trip into the depths of Anthropic's API

karashiiro.leaflet.pub/3m4gf7geefs2l

The New Calculus of AI-based Coding A collection of posts by Joe Magerramov

blog.joemag.dev/2025/10/the-...

Interesting, I see — Whereas the others are more short-termist and beholden to shareholders?

I am curious: Do the others have business models that genuinely couldn’t support higher wages? Or it’s more… a choice?

Well done!

By the 70s/80s, virtually every new or modernised primary school was built with a multipurpose hall that doubled as dining room, assembly hall, and PE space — and the apparatus became a standard feature.

• Firms like Nissen, Huck, and Universal Services supplied standard modular wall-mounted climbing frames, foldable for hall space, with attachable ladders, ropes, and beams.
• Local Education Authorities (LEAs) often bulk-purchased these to equip new or refurbished schools.

Mine (in the 90’s) had this. Summarised from ChatGPT:

• The govt’s School Building Programme (and later DES) building standards included specifications for gymnasia and approved apparatus.