Cover Photo for What's actually changed: building software with AI when everyone feels overwhelmed

What's actually changed: building software with AI when everyone feels overwhelmed

Primary Photo for Madison Artifical Intelligence group

Hosted by

Madison Artifical Intelligence group

In-Person

Address available to attendees

Online

Link available to attendees

We missed you this time around!

AI is changing faster than almost anyone can track. Not the people using AI every day, not the people paying for it, not the people deciding whether to use it at all, and not the people building the architectures. The volume of news is unmanageable, partly because a lot of it is now being written by the AIs themselves. Half of this talk is about what deserves your attention. The other half is Bob.

Software engineering used to have a guy named Bob. Bob's job was to look at what you were doing, ask why, and tell you to slow the hell down. Bob read the spec before the code, the code before the commit, the commit before the deploy. Bob was annoying. Bob was correct. I miss Bob.

Vibe coding does not have Bob.

I tried to get him back into the room by creating four tools. Duplo is Bob creating the spec from whatever reference material you have. McLoop is Bob running the build while you sleep, with discipline about what gets committed. Orchestra is grumpy Bob fighting LLM slop by putting multiple models in specific roles, because any single LLM can fail spectacularly and Bob does not like that. Vroom is Bob reading what shipped and asking what should have been done differently.

Short live demo of Bob. Where Bob came from. Where Bob breaks. Bring questions, and bring the kind of skepticism the field used to have before it fired Bob.

Bio

Mike Coen is a computer scientist and ML researcher whose work spans software agents, self-supervised learning, AI security, and LLM infrastructure. He earned his S.B., S.M., and Ph.D. from MIT, and received the Sprowls Award for outstanding dissertation in computer science. He was previously on the faculty at the University of Wisconsin–Madison and co-founded several fintech startups in New York and Chicago. His recent work focuses on hardening AI systems and working toward 1,000 useful commits per day on GitHub.

What's actually changed: building software with AI when everyone feels overwhelmed

Primary Photo for Madison Artifical Intelligence group

Hosted by

Madison Artifical Intelligence group

In-Person

Address available to attendees

Online

Link available to attendees

AI is changing faster than almost anyone can track. Not the people using AI every day, not the people paying for it, not the people deciding whether to use it at all, and not the people building the architectures. The volume of news is unmanageable, partly because a lot of it is now being written by the AIs themselves. Half of this talk is about what deserves your attention. The other half is Bob.

Software engineering used to have a guy named Bob. Bob's job was to look at what you were doing, ask why, and tell you to slow the hell down. Bob read the spec before the code, the code before the commit, the commit before the deploy. Bob was annoying. Bob was correct. I miss Bob.

Vibe coding does not have Bob.

I tried to get him back into the room by creating four tools. Duplo is Bob creating the spec from whatever reference material you have. McLoop is Bob running the build while you sleep, with discipline about what gets committed. Orchestra is grumpy Bob fighting LLM slop by putting multiple models in specific roles, because any single LLM can fail spectacularly and Bob does not like that. Vroom is Bob reading what shipped and asking what should have been done differently.

Short live demo of Bob. Where Bob came from. Where Bob breaks. Bring questions, and bring the kind of skepticism the field used to have before it fired Bob.

Bio

Mike Coen is a computer scientist and ML researcher whose work spans software agents, self-supervised learning, AI security, and LLM infrastructure. He earned his S.B., S.M., and Ph.D. from MIT, and received the Sprowls Award for outstanding dissertation in computer science. He was previously on the faculty at the University of Wisconsin–Madison and co-founded several fintech startups in New York and Chicago. His recent work focuses on hardening AI systems and working toward 1,000 useful commits per day on GitHub.