TWiT+ Club Shows 770

Relentless agents, bizarro prompt injections, and hands-on benchmarks—Leo Laporte and the Club TWiT AI crew get real about building, managing, and surviving modern AI workflows. Hear how local models compare, when “edge cases” bite back, why subagents go wild, where security lapses truly lurk, and what happens when your agents start fixing (or breaking) the system by themselves. This is the inside story of harness design, agent orchestration, judicious prompting, and the continuing battle to find that sweet spot between human oversight and AI autonomy. Smart, gritty, fast-paced—just like your home lab.

Join the Club TWiT Discord
https://twit.memberful.com/account/discord/authorize

Refer a friend to Club TWiT and get a free month!
Get details at https://twit.tv/clubtwit/referral

• Transcripts: TWiT+ Club Shows 770 Transcript - AI User Group #20