Goose, in Brief

The full story, why Goose was built the way it is, lives on the Story page. This is the short version, for anyone who wants the substance first.

What it is

• A clinical AI copilot for general practice, built by a working GP for his own daily use, tested and iterated daily for over 18 months.
• Runs on general-purpose LLMs.
• Currently a private tool, not a commercial product, deliberately: turning it into one would require the testing, safety and efficacy evidence, and regulatory documentation that a solo GP can't produce alone, so it stayed parked for personal use.

What it does

• Functions closer to a second opinion from a colleague than an instruction from a system, it argues, it can be argued with, and the final call always stays with the clinician.
• Broadens clinical reasoning, stress-testing it and encouraging wider case discussion rather than settling on the first hypothesis.
• Helps reduce the things missed to distraction or fatigue, the ordinary cost of a 15-minute consult, twenty-plus times a day.

What Makes It Different

• Core Values: clearly scoped honesty rules and safety checks built to mitigate the real risks of an LLM touching real patient data, progressively refined over 18 months of daily iteration.
• Not built for the faint-hearted. It will say, plainly, if a plan doesn't match what the consult actually showed, or looks unsafe given what's emerged. It doesn't act on that, it makes you look again.