This Day in AI Podcast – EP99.34

"Am I Even Needed Anymore? GLM-5, Agentic Loops & AI Productivity Psychosis"
Hosts: Michael Sharkey & Chris Sharkey
Date: February 13, 2026

Overview

In this episode, Mike and Chris dive deep into the latest developments in AI, particularly focusing on open-source models like GLM-5 and the rapidly expanding use of agentic loops in their own projects. Threaded throughout is a candid and often humorous exploration of “AI productivity psychosis”—the sense of anxiety, overload, and existential doubt that comes with working and experimenting at the bleeding edge of artificial intelligence. The hosts juxtapose technical insights with reflections on the personal impact of relentless automation, productivity pressures, and what it means to remain "relevant" as AI systems grow ever more capable.

Key Discussion Points & Insights

1. Audience Surprises & The Community

• (00:11–01:41)
  Mike and Chris open by discussing feedback from their podcast audience, humorously shocked by the revelation that someone "quite high up at NASA" listens to the show.
  • Memorable Quote:

    “My message for these people is, what are you doing listening to this crap for?”
    —Mike (00:55)

2. AI Productivity Psychosis: The Human Cost of Agentic Loops

• (01:55–03:27)
  Both hosts share how the rise of AI agentic work has intensified their sense of burnout and decreased work/life balance.
  • Chris: Describes life as a “downward spiral” due to the drive for constant productivity.
  • Mike: Admits to being so obsessed with productivity tools that he can’t focus on family.

    "Even like important moments in my life when I should be very focused spending time with my wife or children and enjoying those moments, I'm thinking, oh, you know what? I have got telegrams set up now to communicate with SIM theory so I could be just spitting out more agent tasks. That's how sick I have become."
    —Mike (02:42)

  • Insight:
    The “AI productivity psychosis” stems from both the power of agentic AI and the relentless pressure to leverage it.

3. GLM-5 Open Source Model — Technical Deep Dive

• (03:55–11:27)
  The hosts discuss the newly released open-source GLM-5 model:
  • Origin: Chinese lab (Zipu, now Z AI).
  • Specs: Mixture-of-experts architecture, 128K output window, pricing "so cheap it's basically free," and zero U.S. hardware dependency thanks to Huawei Ascend chips.
  • Comparison: Early impressions are favorable; GLM-5 performs on par with leading models (e.g. Claude Opus, OpenAI Codex) for routine tasks.
  • Use Case: Particularly suited for “day to day grunt work” and recurring agentic tasks due to low cost and toolchain compatibility.
  • Quote:

    "There's just certain tasks, I think, with models now, where they're sort of indiscernible once they reach a certain level."
    —Chris (07:53)

  • Caution:
    Still, for serious projects or knowledge work, they suggest established commercial models might remain preferable.

4. The Rise & Tuning of Agentic Loops

• (11:27–15:03)
  A “new model war” is underway, but the key battleground is now agentic tuning.
  • All frontier models (OpenAI Codex, Claude Opus, GLM-5) are being optimized for agentic loop workflows—automated sequences of tasks using tools, code, and memory.
  • Quote:

    "All the models now are just tuned to work in this agentic loop and there's a real emphasis in the way they're trained..."
    —Mike (11:27)

  • Some users notice downsides, as aggressive agentic tuning can disrupt more traditional, turn-by-turn chat interactions.

5. Context Windows, Sub-agents, and Practical Workflows

• (15:53–24:17)
  Discussion of context window sizes (128K, 200K, 1M tokens) and their impact on agentic workflows.
  • Key Insight:
    Well-designed agentic loops render massive windows less necessary, as sub-agents focus on tightly-scoped tasks.
  • Chris:

    "Within that you've got things like skills and memory. So...those are spawned as sub agents with even tighter prompts. They don't need to worry about, for example, the overall goal anymore. They're only really worried about their particular part." (17:01)

  • User Experience:
    The human role is increasingly about managing and validating agent output, testing features, and maintaining the high-level vision.

6. The Paradox of Overload: More Power, More Stress

• (24:41–31:17)
  Hosts reflect on the increasing cognitive burden: with more agentic power comes “paralysis of choice” and the potential for burnout.
  • Mike:

    "The limitations you have is purely your own creativity or your ability to ask for things...At this point, I guess it's like being a multi billionaire where you can basically buy whatever you want and so suddenly you've got this paralysis of choice." (25:38)

  • Chris: Notes that “getting 90%” of a task done in parallel just leaves more unfinished work and context-switching.

7. The 95/5 Problem: Edge Cases & Human Relevance

• (32:09–36:51)
  Even with advanced AI, it’s the last 5%—edge cases, final polish, judgment—that remain stubbornly human.
  • Mike:

    “If it was so smart, why aren't I done now?” (27:02)

  • AI can automate the journey, but users need better tools for reviewing, rolling back, or understanding changes—especially as agentic tasks multiply.

8. The Coming Age of 'Command and Conquer' for AI

• (36:51–38:42)
  Prediction: Modern knowledge work will center around “command and conquer” software that orchestrates, monitors, and helps humans validate AI agent outputs.
  • Mike:

    "The command and conquer software that sits on top of all of these agentic tasks to me, is going to become fundamentally the most critical software in the stack..." (35:47)

  • Legacy tools (Atlassian, Trello, Notion) have an opportunity to re-invent themselves as AI command centers.

9. The Future of Work, Collaboration & Role Redefinition

• (38:42–45:50)
  AI will restructure organizations:
  • Roles may meld (e.g., between Product, Marketing, Sales).
  • The future of work may look like assigning tasks on a JIRA/Trello board, with agents executing them end-to-end.
  • The pace of business and expectations for output will increase, but so will overload.
  • Quote:

    “The lines can totally blur…when the agency loop builds a new feature, say in Sim theory, it could automatically announce it and we wouldn't have to do anything.”
    —Mike (44:01)

10. Ten-Year Horizon & Pace of Change

• (46:35–48:44)
  Despite the hype and rapid iteration for early adopters, real organizational and economic transformation will stretch over years.
  • Mike:

    "You just think how long it takes in an enterprise to adopt a new technology, let alone shift job roles and fundamentally how people work. I think this is like a five to ten year thing…” (47:42)

11. From Collaboration to Adversarial AI

• (49:26–50:43)
  Collaboration becomes less with humans, more with agents. In a humorous turn, Mike describes family disagreements mediated (and escalated!) by dueling AI assistants.
  • Mike:

    "My wife and I genuinely will now over like life decisions and stuff, I'll be like, well, my AI says this and she'll be like, but my assistant says this…We get into like debates and I'm like, you probably use some POVO model. I'm using the latest stuff...It just shows how pervasive this stuff now is." (49:59)

12. Industry News: Safety Team Exodus at Major AI Labs

• (51:09–55:11)
  Multiple recent high-profile departures from OpenAI, Anthropic, and XAI, with some exiting staff citing safety and ethical fears.
  • Mike is highly skeptical, seeing these “safety rants” as personal branding for ex-employees more than signals of imminent AI doom.
  • Quote:

    "To me, my opinion of this is not that like we're at some crazy recursive self improvement thing and we're all going to die soon. I put that around 20 years. But where I think we're really at is that these safety people don't matter. They serve no purpose. And they get sidelined in these organizations because they're just annoying." (53:01)

13. Final Reflections: Existential Angst & Hope

• (56:48–58:13)
  Hosts end on a personal note, voicing both high anxiety and awe at the pace of change.
  • Chris:

    "I just honestly hope that I'm still relevant by next week." (56:48)

  • Mike:

    "I do think a lot of this is just excitement around agentic loops finally working the way they've been promised…we should temper our expectations here." (57:38)

Notable Quotes & Moments

• "AI productivity psychosis… I don't know how healthy it is." (Mike, 03:10)
• "It's almost sick where you're like, you're spending time with your family. Like think of the lost productivity." (Chris, 03:03)
• "The limitations you have is purely your own creativity or your ability to ask for things." (Mike, 25:38)
• "If it was so smart, why aren't I done now?" (Mike, 27:02)
• "I feel like I've turned into one of the... copium meme of…this whole podcast so far being 'I'm still relevant too!'" (Mike, 24:31)
• "I just honestly hope that I'm still relevant by next week." (Chris, 56:48)
• "Is this the end? The safety guy's becoming a poet, my friend…" (Mike, 59:41 song lyrics)

Important Timestamps

• 01:55 — Introduction to the productivity “psychosis”
• 03:55 — GLM-5 Model deep dive
• 07:53 — GLM-5 performance and parity with other models
• 11:27 — Agentic loops and new model tuning
• 17:01 — Context management with sub-agents
• 24:41 — The ‘paralysis of choice’ and productivity overload
• 27:02 — Existential doubt: human value vs. AI progress
• 36:51 — Emergence of agent orchestration as critical software
• 38:42 — Redefining organizational roles/structures
• 44:01 — Blurring of roles (Product/Marketing via AI)
• 51:09 — Safety team departures at big AI labs
• 56:48 — Final existential reflections
• 59:41 — “Is this the End?” podcast song lyric moment

Tone & Style

In classic Sharkey Brothers fashion, the show is a blend of self-deprecating humor, thoughtful observation, technical enthusiasm, and a dash of existential angst. The hosts maintain a breezy, bantering rapport throughout, even as they touch on business, personal psychology, and the very future of human work.

Conclusion

This episode is a snapshot of 2026’s AI moment: new models and agentic workflows are transforming how even “average” tech enthusiasts work, but not without spiraling impacts on stress, job definition, and the existential meaning of being “needed” at all. For those riding the wave (or just trying to keep up), Mike and Chris offer camaraderie, technical realism, and permission to laugh (and worry) about the coming future—one agentic loop at a time.