Notes from inside the industry.
Thoughts on AI in chemical distribution: what's working, what to skip, and how operators are putting it to use.
Latest posts.
-
May 29, 2026Microsoft built its whole AI bet on ChatGPT. Now it is quietly leaning on Claude.
Microsoft built its whole Copilot bet on OpenAI, then quietly shipped Copilot Cowork on Anthropic's Claude after putting $15 billion into the company last fall. Models are becoming a commodity, the real value sits in your data, and the distributors with clean pricing and customer data will be ready when the agents show up.
→ -
May 26, 2026Most companies I talk to have signed up for Claude, but their teams are not using it.
Most companies with a Claude license never train their people on it, so the tool sits unused. I offer training and hands-on workshops built on chemical data so teams get a return from day one.
→ -
May 25, 2026Anthropic went from $1B annual run rate to $30B in 15 months. OpenAI is at $24B.
Anthropic's annual run rate went from $1B to $30B in 15 months, and eight of the Fortune 10 are now Claude customers, even though most operators still picture ChatGPT when they think "AI." Brand recognition is a lousy proxy for which model is actually built for your work.
→ -
May 23, 2026Anthropic just signed a 300 MW compute deal with SpaceX. The bottleneck in AI moved from model quality to electricity.
Anthropic signed a 300 MW compute deal with SpaceX for power from its Colossus 1 data center in Memphis, enough for roughly 250,000 homes. The AI race has shifted from who has the smartest model to who can plug in, and that changes what to watch when picking a vendor.
→ -
May 23, 2026I counted 60 AI tools in one infographic this week. Most companies will try a few and go nowhere.
A friend sent me a chart of 60 AI tools across 12 categories, and most companies I work with are piloting a few without ever getting one fully working. My advice is to pick one tool, tie it to one costly workflow, and give it 90 days before adding more.
→ - May 22, 2026
SpaceX builds rockets, satellites, and now AI compute. Anthropic (Claude) is the first big customer.
SpaceX is now selling AI compute from two massive new data centers, and Anthropic is one of its first big customers. When a company known for landing rockets on barges gets into compute, it shows how serious, and expensive, the AI infrastructure race has become.
→ - May 21, 2026
Gemini just copied Claude's pricing model. That makes ChatGPT the best value in compute. It might also prevent Anthropic from raising their price.
Google switched Gemini to compute-based usage limits, matching what Anthropic already does with token consumption on Claude's Pro and Max plans. Claude Code and Cowork have been underpriced for what they deliver, and Google's move (matching Anthropic's model, without publishing the formula) probably buys Anthropic more time before they have to raise the price.
→ -
May 21, 2026You still have a CSR retyping POs into your ERP. Datacor has an agent that ends that.
Datacor's Winter 2026 release reads a customer PDF and writes the sales order straight into the ERP, no template or OCR required. That's the silicon workforce arriving at distributors, an agent doing a job a person used to do.
→ -
May 20, 2026Two acronyms decide who wins online now. Most websites only know one of them.
SEO gets you ranked in Google's results. GEO gets you cited inside the answer ChatGPT, Google AI Overviews, or Copilot writes back to the user, and most sites are only built for the first one.
→ - May 19, 2026
Builder.ai raised over $450M as an AI company. The AI was 700 engineers in India.
Builder.ai raised over $450 million pitching AI that built apps, when the work was actually done by about 700 engineers in India writing code by hand and told to work UK hours so it looked automated. It's still the best test for any AI vendor pitch, ask exactly where the model decides, what data trained it, and how you validate the results.
→ - May 19, 2026
We design warehouses for people who walk or drive a fork lift. Amazon stopped doing that years ago.
A tour of an Amazon fulfillment center got me asking why distributors still move people toward product when the product could move instead. Hazmat keeps robots out of our warehouses, and there are still lessons in how Amazon rebuilt the question from scratch.
→ - May 16, 2026
Cuts will happen, but don’t cut too deep.
A quote worth sitting with on the GM, Ford, and Stellantis job cuts. Cut your workforce beyond what AI can actually replace and you lose productivity, institutional knowledge, and your best people. Fail to rethink how work gets done and your competitors pass you by.
→