Intelligence Logs

Zero-Click Marketing (Part 4): AEC Proof of Work — The Portfolio Is the Price of Entry. The Tech Receipt Validates the Parking Ticket
Why a small Pacific Northwest practice spent 2026 shipping AI for itself — and how half the work is now the hidden team behind other people's drawings. 50% design firm, 50% shadow fulfillment and BIM/AI retooling.

Zero-Click Marketing (Part 3): The Chef Can Discuss the Dish — What Agent2Agent Uncovered in Our Own Stack
We taught our MCP storefront to converse over Google’s Agent2Agent protocol, then ran 215+ hostile probes. Seven worms underneath: a data plane open to the public key, an ownership cookie minted on demand, a signing key shipped in every bundle, and a fix that silently gagged our own agents.

Zero-Click Marketing (Part 2): The Front Door Is an API — What Happened When We Attacked Our Own MCP Storefront
We attacked our own agentic MCP storefront: the findings, the fixes, the production outage our own hardening caused, and why an endpoint with a mind on the serving side is a different class of target.

Zero-Click Marketing: How Axoworks Built an MCP Connector for the Agents Who Never Click
If AI agents can just read your chunked content, why would anyone need a live LLM? Our answer cost one evening, three tools, and zero new infrastructure.

We Added RAG to DSH. The Vector Database Was the Last Decision We Made.
On a local-first RAG stack, the binding constraint isn’t your embedding model or vector DB — it’s VRAM context budget. How we measured GGUF metadata, avoided CUDA crashes, and picked SQLite.

The Machine Wears the Harness Now. You Just Talk.
DeepSeek built a coding tool. Because everything in it is a plugin, it became a structured harness that puts deterministic rails on stochastic AI — and hands the human an interface that needs no training and evolves with them.

Buy the Model or Rent the Intelligence? The September 2026 Math for a Four-Person Architecture Firm
We priced an Apple cluster against an API bill — including the intern's dead air — and the answer wasn't close. A fact-checked field guide for the smallest 60% of firms.

Death of the Dashboard: The Quota Tax, the Owned Harness, and the Benchmark That Proved It
We benchmarked QwenWork’s pitch against a harness we own, did the math on subscription quotas, and published our own audit panel’s blind spot.

The $1 Property Report: Turning DeepSeek Harness Into a Senior Zoning Researcher
Realtors, developers, and architects — the pre-construction research department you couldn't afford is now an open-source AI preset for DeepSeek Harness.

The Referee in the Pit: How We Built an AI Debate Team to Catch Our Own Models Lying
How Axoworks built an AI debate team — Advocate, Prosecutor, Honey Badger, and Referee — to cross-examine models, catch hallucinations, and enforce factual validation.

revit-tools: A Python MCP Server for Autodesk Revit 2027
An AxoWorks Technical Review: Connecting AI chat clients to Autodesk Revit 2027 via a zero-dependency Python MCP server with deterministic geometry audits.

Why General-Purpose AI Fails at AEC Pre-Construction — and the Verification Stack That Fixes It
An AxoWorks Technical Framework Report: failure modes, 2026 frontier model threat assessment, and the 5-layer verification stack that makes AI output trustable enough to build on.

The Lineup: Which AI You Bringing to the Heist?
Ever watched Ocean’s 11? The real con happened the moment you picked your model. A breakdown of the profiles, tells, and verdicts of the AI accomplices in your stack.

The Three Dog Theory: Is Your Office a Dog Shelter?
Most offices are AI dog shelters — puppies that hallucinate and confused shepherds waiting for orders. The winners build Elite K-9 units. Which one are you running?

The Three Dog Theory: Is Your Office a Dog Shelter? (August 2026 Update)
The original theory, stress-tested over a year of real-world deployment. What actually changed when the first true consumer AI agents hit the market.

AI Fight Club: Why Tiny Teams Beat Corporate AI
How a tiny team running a private arena of 20-30 open-source LLMs routinely outperforms corporate teams betting everything on one neutered frontier model.

The Axoworks Logs: Surviving the Conversational Web
We deleted our brochure and built an AI terminal. Hacker News attacked it, crawler swarms probed it. A build log from the conversational web, Dec 2025 - Mar 2026.

The $0.74 Website: How Axoworks Replaced Its Storefront with an AI Agent
We killed our Wix brochure and launched The Concierge — an agentic AI storefront that survived the Hacker News front page for $0.74 in AI hosting costs. The story of our no-browse website experiment.

From Portfolio to Agent: How Axoworks Built One of the Web's First Agentic AI Websites
A full technical analysis of the Concierge — the distributed Brain-Hands-Voice loop, the Eager RAG trade-off, the SEO paradox, the guerrilla logs.axoworks.com bridge, and the engineering traps that came with it.

LLMs for Real Estate Feasibility: Your AI Ace Up the Sleeve
How LLMs with RAG crack zoning codes in seconds, spin dozens of pro forma scenarios, and sniff out tax credits — de-risking feasibility studies for developers.

Why Open Source Local LLMs Are Your Privacy Superstars
Cloud AI tools slurp up your client data. Open-source local LLMs keep everything on your own hardware — total control, offline power, no subscription bleed.

The Magic Pill: Generative AI for Design Presentations
Generative AI cuts design presentation timelines from weeks to minutes — and a local setup keeps your client work locked down while you do it.

Code Chaos to Creative Clarity: LLMs + RAG vs. Building Codes
Building codes read like alien legal diaries. LLMs with RAG translate them in seconds and flag liability before it becomes a lawsuit.

Three Ways an AI Lies to You (And Only One of Them Is Censorship)
An adversarial benchmark across frontier language models reveals three structurally distinct ways AI fails to tell the truth while sounding authoritative.

Coercing AI Compliance: Structural Rails for Consistent Multi-View Architectural Visualization
Case study: four photorealistic, material-consistent renders of a mountain lodge from hand sketches in 40 hours — semantic color-coding, LoRA, ControlNet, and SAM on an RTX 6000.

Coercing AI Compliance: A 4-Layer Structural Rails System for Consistent Multi-View Architectural Visualization
How Axoworks produced four viewpoint-consistent, photorealistic exterior renders of a luxury mountain lodge from hand sketches alone — no BIM model for rendering, no 3D geometry export.

The Hybrid Delivery Workflow: You Control the Design, We Wrestle the Tech
Keep authoring in AutoCAD. AxoWorks runs Revit + AI in the background — 3D coordination, clash detection, and instant AI-enhanced presentations.