A recent Mosh Bari video on: Anthropic Just Told Everyone to Stop Building AI Agents… Here’s Why. This post highlights the core topic and why it matters, with the full discussion in the embedded video.
Why it matters: As more teams ship agent workflows, reliability and safety depend less on clever prompting and more on engineering: data contracts, tool boundaries, and observability. Good system design is what turns demos into something you can run daily.
Singularity Soup Take: The agent boom is going to create a lot of fragile, quietly-dangerous automation; the winners will be the teams that treat agents like software—with tests, permissions, and monitoring—rather than like a chat trick.
Watch on YouTube — Mosh Bari