Publications
- Private but Credible: Domain, Hosting and Mail That Google and AI Search Still Trust
How to keep your personal details out of public records while staying the kind of site Google ranks and AI engines cite — across domain registration, hosting and email, targeting North America and Japan. The rule that resolves the tension: redact the person, publish the entity, and never fabricate one.
- Buying Tokens for Autonomous Agents: Who Sells API Credit, and What It Costs
The companion articles covered subscriptions, which an autonomous agent cannot use. This one covers metered API tokens — what the big-tech vendors charge, what you can buy from outside the US, and what the Chinese labs charge, plus the spend tiers and token amplification that bite long before the per-token price does.
- Big Tech AI Subscriptions: What Everyone Actually Charges
Every major lab and hyperscaler that trains its own models and sells a consumer subscription — OpenAI, Anthropic, Google, Microsoft, Meta, xAI, Amazon and Apple. Current prices across all tiers, what each is genuinely best at for code and data work, and where each one is weak.
- Chinese AI Services: Which Ones You Can Actually Sign Up For, and What They Offer
Most Chinese AI services are reachable from outside China with a plain email address, and several lead their category outright. Which ones work internationally, what they cost, what they're best at — and the four caveats that decide whether you should use them.
- If the Big Labs Block You: Every Alternative That Trains Its Own Models
One list of every independent developer that trains its own models and sells self-serve access — text and code, images, video. No big tech, no Chinese developers, no research projects you'd have to host yourself. Twelve companies, with prices and where each one is weaker than the leaders.
- Fifteen Free Models for a 128GB Strix Halo Notebook: A Working Loadout for Text, Code, Images and Speech
What actually runs well on a Ryzen AI Max+ 395 with 128GB of unified memory — fifteen freely downloadable models chosen by task, with download sizes, measured or estimated tokens/sec on this specific chip, licence traps, and the counterintuitive rule that decides all of it: on this hardware, bigger-but-sparser beats smaller-but-dense.
- Self-Hosting the Frontier on Multi-GPU Servers: 10 Free-to-Download Models and 5 You Pay For (September 2026)
A sizing table for professional self-hosted inference — the most capable open-weight models you can download today, with parameter counts, checkpoint sizes, the GPU memory each needs resident, and what each one is actually best at — plus five frontier models you can run on your own hardware only under a commercial licence.
- AI & VR Hardware Architecture: Portable 128GB Systems for Local LLMs and Wireless VR
A working comparison of four portable/compact hardware architectures for running 70B+ local AI models alongside high-fidelity wireless VR — corrected against current vendor specs, with CUDA-development suitability and a recommendation.
- Sourcing the Stablecoin Beat: Free Feeds, Paid Data, and What They Cost
One analyst's working list of free and paid sources for tracking the stablecoin market broadly, with price ranges for the paid tier.
- AI Services Directory: Chat, Code, Images, Video & Deep Research
A curated directory of AI chat and generation services, sorted by use case and free vs. paid, plus a section on aggregators. Refreshed September 2026 — Sora removed, agentic coding CLIs and the current video generation field added.
- First publication
Placeholder publication.