Thoughts & Perspectives
Building in public
Haiku vs Sonnet (2026): I Ran Both on Real Business Jobs. Haiku Was 12 Times Cheaper, Not 20
Claude Haiku 5.5 lists at a twentieth of Sonnet 5.5's price. On a measured test of email routing, invoice extraction and pricing math, it scored 154 of 156 to Sonnet's 156 and cost 12 times less, because it writes more output tokens. When each model is the right call, as of October 2026.
Grok vs ChatGPT (2026): Prices, Limits and Team Plans After The X Subscription
Grok and ChatGPT both start at $8 now, but they split right after: ChatGPT Plus is $20, while Grok's main tier is $30 and bundles Cursor and X Premium. The two ladders side by side, how each one meters usage, what a 10-person team pays, and which to buy, from both vendors' own pricing pages as of October 16, 2026.
ChatGPT for Marketing (2026): The Setup for a Small Team, and ChatGPT as the Place Customers Find You
"ChatGPT for marketing" means two things now. It's a tool: a project that holds your brand voice, scheduled tasks that draft the weekly work, and image generation that eats your usage 3 to 5 times faster. And it's a channel: ads that only Free and Go users see, plus the organic answers your B2B buyers actually read. The setup and the cost of each, as of October 2026.
Manus Pricing (2026): Free, Pro $20 or $40, Team $20 a Seat, and What a Credit Buys
Manus is free with 300 daily credits (capped at 1,500 a month, Lite model only), Pro starts at $20 a month for 4,000 credits or $40 for 8,000, and Team starts at $20 a seat. What a credit costs, why the $40 tier isn't cheaper per credit, and what a team pays, from Manus's own help center as of October 2026.
AI in Auditing (2026): What a Small CPA Firm's Audit Team Can Hand AI, and What the Standards Say About the Evidence
How a small CPA firm should use AI in its audit practice: let it draft PBC lists, planning memos and workpaper narratives, treat anything it produces as information you still have to test, and keep judgments and the opinion with the engagement team. With the SAS 142 and PCAOB rules side by side.
AI for Banks (2026): What a Community Bank or Credit Union Can Hand AI, What Stays Human, and How to Vet the Vendor
How a community bank or credit union should start with AI: draft procedures, member letters and credit memo narratives, keep credit decisions and BSA work human, sort every bank record before it goes near a tool, and run the AI vendor through third-party risk management like any other vendor.
Gemini Gems Are Becoming Skills: What Changes and When (2026)
Gemini Gems are saved custom instructions in Google's Gemini app, and Google is replacing them with skills. Skills started rolling out on October 13, 2026. Personal accounts lose Gems starting in November; Workspace business accounts no sooner than March 1, 2027, when leftover Gems turn into inactive draft skills. What changes, the dates, and what to do with the Gems you already have.
Is Perplexity Pro Worth It? Only If It Replaces Something You Already Pay For
Perplexity Pro is $17 a month billed annually. It's worth it if your work is research with sources and you'd drop another $20 subscription for it. It's a bad buy stacked on top of ChatGPT Plus or Claude Pro, because Perplexity answers with those same model families. The overlap test, what Free leaves out, and who should skip it, as of October 2026.
Perplexity Pricing (2026): Free, Pro $17, Max $167, and $40 Enterprise Seats
Perplexity costs $0 for Free, $17 a month for Pro and $167 a month for Max when billed annually (Max is $200 month to month). Enterprise Pro is $40 a seat, Enterprise Max $325. Every tier, the seat math for a team, and the billing rules that catch admins out, from Perplexity's own pages as of October 2026.
Claude for Financial Services (2026): What Anthropic Actually Sells, and What a Small Firm Should Buy Instead
Claude for Financial Services is Anthropic's package for banks, insurers and asset managers: models, data connectors, Excel and PowerPoint, and a sales team. Here's what's in it, what the new advisor plugin does, and the plan math for a 6-to-25-person firm.
AI Training for Employees (2026): What to Teach Whom, in What Order, and What a 50-Person Rollout Really Costs
A practical plan for training staff on AI: a 90-minute baseline for everyone, extra hours for managers and a few champions, the staff-hours math for a 50-person company, and what the EU AI Act's literacy article actually says now.
HIPAA Compliant AI (2026): Which Claude, ChatGPT, Gemini and Copilot Plans Get a BAA, and What It Costs
No AI model is HIPAA compliant; a signed Business Associate Agreement is. As of October 2026 all four big vendors sign one, but each draws the line at a different plan, and two of them exclude the business plan a practice would naturally buy. One table, read from each vendor's own pages this week, plus what the covered path costs a 10-person practice.
Ready to scale your product vision?
I help high-growth startups bridge the gap between technical complexity and business revenue.
Book a Strategic Call