AI
Alien Intelligence on Your Desk: Why AEC Can't Benchmark the AI It Trusts
Melanie Mitchell calls LLMs 'alien intelligence' we can't measure. What that means for any architecture office wiring an unauditable model into its workflow.
Karpathy's Homepage Is a Dependency Graph — the AI-for-AEC Lesson Hiding in a CV
Andrej Karpathy's karpathy.ai reads like a dependency graph - and it's the from-scratch lesson every AEC office adopting LLMs in 2026 needs before wiring one in.
The CNN, Read Like a Beam Grid: One Filter, Checked by Hand
A foundation guide to the convolutional neural network for architects: the checkable arithmetic, the Hubel-to-ResNet lineage, and how to audit a filter's output.
Grok 4.6 Ships Into Your Toolchain by Default — Write the Exit Clause First
Grok 4.6 lands in Cursor, Copilot and Bedrock at $2/$6 per million tokens. Pin the model string, log every call, and govern the default before it governs you.
The Provenance Stack: A PAZ Office SOP for AI Watermarks
Anthropic marked Claude's text on 2 Aug 2026; a stripper followed in ten days. The four-layer provenance system, and a PAZ office SOP for handling it.
Seven Numbers, No Memory, 96.5%: The 14-Byte Brain That Shames Your Render Farm
MINIMIO's 14-byte neural net solves 96.5% of unseen mazes with no memory — and quietly teaches AEC teams when reactive beats heavy tooling.
A Fields medallist quit maths for AI safety — the lesson is where the gate sits
Fields medallist Jacob Tsimerman left maths for OpenAI's AI-safety work. The lesson for AEC offices: put the guarantee in the wiring, not in trust.
DeepSeek-V4 ships a million-token context at a tenth the KV cache
DeepSeek-V4's preview cuts KV cache to 10% and FLOPs to 27% at a million tokens — the infrastructure move that makes long-context AI affordable for AEC offices.
Kimi K3 Hits the Studio: a 2.8-Trillion Open Model That Reads Your Screenshots
Moonshot's Kimi K3 is the first open-weights 3-trillion-class model, with 1M-token context and native vision that names CAD. A PAZ hands-on for BIM desks.
Cache-augmented generation: the cheapest token is the one you never re-read
How CAG caches a frozen corpus to cut RAG cost — the O(n²) attention bottleneck, Redis TTL traps, and where a Swiss BIM studio should actually reach for it.
