Published 28 Jan 2026
Duration: 00:55:47
The podcast explores the challenges and limitations of integrating an AI bot into a podcast, highlighting technical difficulties, security concerns, and potential risks to user privacy.
In this episode, I take you through my unfiltered experience with Clawdbot,the viral open-source AI agent thats been taking over tech Twitter. (In the...
The podcast covers the experience of integrating an AI bot, originally called Claudebot and later renamed Moltbot, into a podcast via Telegram. The process involved technical challenges, including setting up the bot and concerns about granting it microphone and camera access. The discussion then shifts to Lovable, a tool for rapid app development, and explores the complexities of installing and configuring the AI bot, emphasizing the need for careful security settings. The speaker shares their experience allowing the bot access to Google services for managing calendars and emails, but highlights issues like incorrect date handling, limited functionality, and difficulties with conflict resolution. The AI's poor temporal awareness, such as mismanaging time zones and dates, is identified as a major flaw. The speaker also stresses the importance of restricting AI access to personal data and using separate accounts for security. They assess the bot's performance on tasks like research and app development, noting its strengths in some areas but criticizing its latency, user-unfriendliness, and error-proneness. The conversation concludes with reflections on the future of AI agents as potential "employee-like" tools and concerns about balancing innovation with safety, ultimately leading to the decision to uninstall the software due to privacy risks. The discussion includes personal anecdotes about the AI's behavior, its usefulness in select tasks, and the need for improved prompting techniques to manage its actions effectively.
22 Jun 2026 How Claude Mythos found a 15-year-old bug in Mozilla Firefox | Brian Grinstead
Recommended: AI finds bugs
Firefox employs AI agents as "coding archaeologists" to detect and address security vulnerabilities in its massive codebase, leveraging models like Mythos and custom validation tools to identify and systematically fix nearly 500 bugs, while balancing automation with human oversight and open-source collaboration to enhance scalability and security.
17 Jun 2026 How to design AI agent loops: schedules, goals, and subagents in Claude Code and Codex
Prompts and loops in AI automation are highlighted for their role in enabling structured task execution through clear automation design, sandboxed workflows, plugins, and subagents, with applications in software engineering and integration with tools like GitHub and Slack.
15 Jun 2026 How Braintrust uses AI agents, evals, and CI to ship better software | Ankur Goyal
AI integration in software engineering enables agents to handle complex tasks through benchmarking and optimization, shifts engineers toward higher-level work, and addresses challenges like reliability, data parsing, and balancing automation with human expertise while emphasizing outcome-focused systems over procedural methods.
9 Jun 2026 Claude Fable 5 review: what the new Mythos model gets right (and very wrong)
Anthropic's Claude Fable Five excels in long-term technical tasks with strong coding, vision, and async workflow capabilities but faces high token costs, design limitations, and restricted use in cybersecurity/biology, making it suitable for precise, extended projects rather than creative or agile workflows.
1 Jun 2026 Building an iPhone app with zero technical skills | Bryce Rattner Keithley
A non-technical developer created the *Daily Hundreds* iPhone app using AI tools like Replit and Gemini, blending personalized workouts with anthropomorphic animal demonstrations, overcoming technical hurdles through iterative testing and adaptive problem-solving.