It’s the January 29 AI Update for the AI x Higher Ed podcast—and the theme is simple: AI is moving from “chat” to “do.”Stephan breaks down Moltbot (formerly “Claudebot”)—an open-source, agentic assistant you can run on your own machine or a private server, complete with persistent memory, recurring tasks, and messaging-app workflows (Telegram/WhatsApp). Then the conversation turns to Claude Co-Work, where “no-code” agentic automation can organize files, write scripts, download and summarize reading sets, and even draft emails through your browser—hinting at a near future of parallel AI coworkers running tasks simultaneously.Finally, the hosts unpack Dario Amodei’s new long-form essay, “The Adolescence of Technology”: timelines for “powerful AI,” why speed and generalizability make this wave different, and what higher ed should be thinking about now—jobs, governance, disruption, and the ethics of adoption.If you’re experimenting with agents (Moltbot, Claude Co-Work, Claude Code), they want to hear what you’re trying—and how it might translate into classroom workflows.Chapters00:04 January 29 AI Update kickoff — what a month already00:16 Substack format shift: weekly roundup + what we’ll focus on today01:50 Moltbot (formerly “Claudebot”): why the name changed02:10 The “space lobster” mascot + sci-fi roots (Doctor Who + Accelerando)03:47 What makes Moltbot different: not a chatbot—an agent on your device04:22 The download spike + quick setup (AWS + API key)05:26 Why people get hooked: persistent memory, recurring tasks, proactive pings06:01 The interface shift: running through Telegram/WhatsApp like a “coworker”07:36 Naming the bot “HAL” + what it reveals about personality + UX08:41 Why Moltbot matters: a stepping stone to mainstream personal agents09:44 Claude Co-Work in practice: downloading a whole reading set as PDFs10:52 When “print to PDF” fails: Claude writes a script and adapts11:38 Paywalls + workarounds + speedups via URL pattern learning12:04 Turning a folder of PDFs into a structured Substack post13:21 What Claude Co-Work is (and isn’t): no-code, paid accounts, folder control14:34 Real workflow demo: drafting/sending an Outlook email via the browser16:42 Why slow now doesn’t matter: the future is voice + instant execution17:49 Under the hood: screenshots → interpretation → action (and why that’s slower)18:26 The “real” power: running many agents in parallel19:18 Using Co-Work for critique: research paper evaluation + course proposal automation20:45 Paperwork killer: guidelines → proposals → spreadsheet requirements check21:52 Finding the sweet spot: multi-agent orchestration for higher ed work23:33 Dario Amodei’s new essay: what it is and why it matters24:30 Timeline talk: “powerful AI” as early as end of 202625:13 Long-horizon tasks + millions of copies: why this scales differently26:18 Risk landscape: cyber, bio, surveillance, power concentration27:50 The speed problem + real job displacement signals28:31 Generalizability: why this tech isn’t like past automation waves31:00 “Character and determination”: governance, norms, and what we choose33:24 Why take Amodei seriously: incentives + proximity to capability34:33 Recursive self-improvement: AI shaping its own “constitution”37:03 Closing: stop debating “if” and start planning “how”38:05 Call for listener experiments: agents + classroom applications#aixhigheredpodcast #HigherEd #ArtificialIntelligence #AIUpdates #AgenticAI #AIAgents #ClaudeAI #Anthropic #EdTech #FutureOfWork #AcademicWorkflow #AIProductivity #Automation #DigitalAssistants #AIandEducation