The Breakeven Brothers are finally back! After another hiatus filled with cross-continental travel, surprise kitten fostering, and a whole lot of life, Bradley and Bennett jump right back into the deep end. We kick things off with stories from Brad's trip to Korea—navigating rainy weddings and constant hotel hopping—and Bennett's unexpected turn as a foster dad to a tiny kitten that wandered into his backyard. These adventures set the stage for a much larger conversation about how quickly things can change, especially in the world of AI.
Once we're settled, we dive into a wild new technique from DeepSeek that turns text into images just to... turn it back into text? It sounds crazy, but this 10x compression trick could fundamentally solve one of AI's biggest frustrations: why our agents seem to get dumber the longer we talk to them. This sparks a huge debate about OpenAI's new browser, ChatGPT Atlas. Is it a revolutionary agent that will plan our trips and buy our groceries, or a massive security risk waiting to happen? Brad also gives us a firsthand report from an exclusive "Claude Code Anonymous" meetup in San Francisco, sharing the secrets behind a 700-line master prompt that gets near-perfect results. Tune in for a packed episode covering everything from AI walled gardens to the future of how we browse the web.
Chapters:
00:00 Back from Hiatus & Personal Updates
04:11 DeepSeek's Text-to-Image Compression
13:15 OpenAI's New Agentic Browser: ChatGPT Atlas
20:38 Security Concerns with Agentic Browsers
33:20 Takeaways from Claude Code Anonymous in SF
40:16 Bookmarks: Meta vs. AI & Cursor's Plan Mode
43:54 Conclusion & wrap up
Links:
- Meta Banning Rival AI Chatbots on WhatsApp
- Cursor's New "Plan Mode"
#AI #TechPodcast #OpenAI #ChatGPTAtlas #DeepSeek #ClaudeAI #AICoding #ContextWindow #AIAgents #BrowserWars