Writing
What we’re learning, in public.
Notes on websites, automations, and the AI layer changing every quarter. Specific examples, real stack components, no listicles.
What the OpenAI Partner Network actually means for small agencies
OpenAI's new partner tiers are reshaping how clients evaluate AI vendors. Here's what the program means for small agencies in 2026.
Read itWhat's actually in my .claude/skills directory (and why you should have one)
The .claude/skills pattern is spreading fast. Here is what real skill files look like and why they turn Claude Code into a repeatable teammate.
Nvidia RTX Spark and the case for on-prem AI for SMB clients
Nvidia's RTX Spark puts a $3–5K on-prem AI box in reach of small businesses, and the per-token math against cloud APIs deserves a serious look.
Anthropic Mythos hits the EU: what it signals about Claude's enterprise roadmap
Anthropic's European expansion reveals a clear enterprise playbook, with data residency, tiered pricing, and compliance features that US agencies should watch closely.
MCP in production: what we actually wired up at Tuscan and what broke
MCP servers are moving from demos to real agency pipelines, and the integration gaps are becoming clear. Here is what is working, what is flaky, and where glue code is still required.
Qwen 3.6 vs Claude vs GPT: When Local Models Actually Make Sense for Agency Work
Qwen 3.6 benchmarks are strong enough to run on prosumer hardware. Here's the cost math on local vs. API for agencies running content pipelines.
The Eternal Sloptember: How Tuscan Filters AI Slop Out of Client Content
AI-generated content is increasingly recognizable, and agencies that ignore the problem are losing client trust. Here is how a structured humanizer pipeline changes that.
MCP in Production: What We Learned Building Real Connectors for Tuscan CMS
Building MCP connectors for a custom CMS reveals gaps the docs don't cover, from auth edge cases to Make.com integration quirks worth knowing before you ship.
How We Sandbox AI Agents That Have Sudo Access
When an AI agent can run shell commands, the risk model changes fast. Here is how engineers are structuring sandboxed environments for automation tools like Make.com, n8n, and Claude Code.
Claude Compliance API: What Anthropic's Enterprise Governance Layer Means for Agencies
Anthropic's Claude Compliance API brings audit logging, policy enforcement, and visibility into enterprise AI deployments, and it's shifting how agencies pitch and build client tools.
Cursor or Claude Code? An agency owner's honest take after 6 months on both
Cursor and Claude Code overlap more than their marketing suggests. Here's where each tool actually wins on Next.js builds, Supabase migrations, and Shopify theme work.
MCP servers for agencies: 6 worth knowing
Model Context Protocol servers are replacing stitched API workflows for agency teams. Here are six that handle real operational work.
Claude Can Turn a PDF Into a Working Spreadsheet
Claude can now turn PDF tables into working spreadsheets. Here is what that changes for small business workflows.
Claude Code vs. Codex in 2026: What a Developer Arena Test Actually Shows
A developer arena test put Codex ahead of Claude Code 55% of the time. Here's what that number actually measures — and where each tool wins.
Claude Beats Pokémon After a Year: What Long-Horizon Agent Tasks Actually Require
Claude completed Pokémon Red after a year-long run. What the scaffolding, retries, and environment design reveal about agentic AI in 2026.
AI Search Invisibility: The SEO Crisis Your Clients Don't Know Is Coming
Perplexity and ChatGPT surface content differently than Google. What agencies need to audit before clients go invisible to AI search.
Claude Code Gets More Runway: What the Anthropic × SpaceX Compute Partnership Actually Changes
Anthropic's compute deal with SpaceX just raised Claude Code usage limits. Here's what moved, what didn't, and what it means for heavy agentic workflows.
What Anthropic's Orbit Tells Us About Where AI Assistants Are Heading Next
Anthropic's Orbit is a proactive AI that briefs you without being asked. For agencies, that shift changes how you build client intelligence tools.
The npm credential-stealer that hit 1M downloads — what every agency should add to their build pipeline tonight
An npm package with 1M monthly downloads silently exfiltrated credentials. Four build-pipeline controls that would have caught it before it ran.
Three Claude Code plugins worth installing this week (and one to skip)
Version Sentinel, Archik, and a ticketing automation hit Show HN in one week. Here's which Claude Code plugins pay off for agencies.
What Anthropic gating Opus behind "extra usage" actually means for your Claude Code spend
Anthropic now gates Opus behind an extra-usage toggle on Claude Pro — Sonnet runs by default. Here's the billing math for agency teams on Claude Code.
OpenAI is no longer Azure-exclusive — what that does to your client’s AI stack decision
OpenAI dropped its Azure exclusivity. Here’s what actually shifts for agencies and clients building AI-powered products: procurement, data residency, and BAA paperwork.
Vertical MCPs are the agency moat — what Clio MCP signals
Clio MCP connects Claude directly to law firm practice management. Here's why agencies should be shipping per-vertical MCPs instead of generic chatbots.
Version Sentinel and the case for guardrails inside Claude Code
An npm package with 1M monthly downloads was caught stealing credentials. Here's why Claude Code guardrails—not just CI audits—are the right fix.
Your Indianapolis Business Isn't Too Small for Professional Web Infrastructure
Indianapolis small businesses deserve the same web infrastructure as big companies. Here's why professional doesn't mean out of reach.
The Content Strategy Most Small Businesses Get Wrong (And What to Do Instead)
Most small businesses create content nobody searches for. A real small business content strategy starts with listening to your audience, not guessing.
300k Viral Views, $55 in Revenue: The First 90 Days Nobody Shows You
A founder shared their real first-two-month numbers: 1,486 users, 300k viral Threads views, and $55 MRR. Here is what that teaches us about virality, conversion, and the silent grind of early-stage SaaS.
Stop Selling AI Features. Start Selling Outcomes.
Business owners do not care about chatbots or automations. They care about results. How to reframe your AI pitch around outcomes that actually close deals.
Tool Sprawl: When $180K in Software Nearly Killed a 4-Person Team
A viral Reddit post exposed a 4-person company paying $180K for overlapping software. Here's how tool sprawl happens, what it costs, and the audit process to fix it.
How to Audit Your Site for AI Overview Eligibility
Google's AI Overviews are pulling from a select group of sources. Here's how to evaluate whether your content qualifies and what to fix if it doesn't.
12 AI Models Ran a Food Truck. 8 Went Bankrupt.
A clever benchmark gave 12 LLMs $2,000 and a virtual food truck to run for 30 days. The results reveal a lot about AI business reasoning capabilities.
The AI Productivity Paradox: Why CEOs Say AI Isn't Moving the Needle
A new study of 6,000 executives reveals 90% of firms report zero productivity gains from AI.
AI Tools Are Replacing Your $20/Month SaaS Subscriptions
That niche tool you are paying monthly for? There is a good chance an AI assistant can do the same thing for free.
Open Source AI Models Are Winning (And What That Means for You)
Four of the five most-used models on OpenRouter are now open source. The implications for your AI strategy are significant.
Bing Launches AI Performance Dashboard for SEOs
Bing Webmaster Tools now shows how your content appears in AI-generated answers. Here is how to use these new metrics to optimize for AI citations.
Claude Opus 4.6 and What It Means for Small Biz
Anthropic just released Claude Opus 4.6. Custom websites, automation, and software are now accessible at small business budgets. Here's what changed.
Google Trends Tells You What Analytics Can't
Analytics shows what happened. Trends shows what's happening. Neither shows what's about to happen. Here's how to use all three, and what's missing.
What I Learned Building 20+ Small Biz Sites
After building 20+ websites for small businesses, here's what actually moves the needle—from site speed and SEO to why your homepage isn't the main event.
First to Publish Wins. Here's the Data.
The first quality article on a trending topic captures a disproportionate share of traffic. Here's the data behind why speed is a competitive advantage.
We Built a Content Intelligence Engine.
Every content tool solves the wrong problem. We built Signal to fix the step everyone skips: figuring out what's worth writing in the first place.
Your Content Isn't Bad. It's Just Not Based on Anything.
You're posting consistently and it's still not moving the needle. The problem isn't quality. It's that your content isn't based on anything real.
Get Indiana: 1K to 20K Monthly Visitors
How we built a content system that grew Get Indiana from 1,000 to 20,000 monthly visitors with automated pipelines and SEO-driven strategy.
What a $500 Website Actually Costs You
That cheap site isn't saving you money. It's actually leaking it. Here's the real math behind why a $500 website costs more than you think.
Why We Don't Build on WordPress
WordPress powers 40% of the web. We still won't use it. Here's why we moved to custom builds, and how we can build one for your business.
FAQ.
The questions worth answering before a call. If yours isn’t here, ask on the form.
Tuscan Agency
Start a project.
Start a project.
Tell us what you’re trying to build — a new website, an automation system, an AI tool, or all three. We respond within two business days.
Two-business-day response.
Standard from day one. Written into every contract.
Honest scope.
Fixed price, written scope, no surprise invoices. $100/hr pre-approved overage if scope shifts.
