Pharma AI
The PMCPA censured Theramex for an unqualified safety claim on LinkedIn. The agency has now clarified that corporate copy carries the same compliance obligations as promotional material. Here is what that means for any pharma team using LLM tools.
Oscar Espinoza · 28 Apr 2026 · 5 min read
Read the post ›
Frontier AI
Opus 4.7 regressed on long-context benchmarks and costs roughly 35% more per request. Anthropic signed a $25B deal with Amazon four days later, citing infrastructure strain. What that means if you run Claude in production.
Oscar Espinoza · 22 Apr 2026 · 5 min read
Read the post ›
Regulated industries
Compliance-heavy sectors have the most structured, repeatable data, perfect for AI. So why are they last to adopt it? And what does tractable actually look like?
Oscar Espinoza · 18 Mar 2026 · 7 min read
Read the post ›
B2B software
Most B2B SaaS fails because it's built before the problem is understood. The services-to-SaaS playbook, use consulting to find a workflow worth productising, then build it, is slower but almost always right.
Oscar Espinoza · 18 Mar 2026 · 6 min read
Read the post ›
Shipping AI
ioomm.app is a live AI product that takes real payments. Here's what building it taught us about the gap between a demo and something that actually runs in production.
Oscar Espinoza · 18 Mar 2026 · 6 min read
Read the post ›
AI governance
Anthropic appointed a pharma veteran to its safety board. The choice signals something specific about how frontier AI labs are thinking about regulated-domain risk — and what that means for enterprise buyers due diligence.
Oscar Espinoza · 25 Apr 2026 · 5 min read
Read the post ›
Coming next
Case studies in progress
We're writing up the first engagements, anonymised, with the trade-offs we actually made. Check back soon.