This Day in AI Podcast

Episode: “2026 Existential Crisis, Claude Code Hype & Is SaaS Dead?” (EP99.30-WIZARDS)

Hosts: Michael Sharkey & Chris Sharkey
Date: January 19, 2026

Episode Overview

Michael and Chris open the first episode of 2026 by confronting the “existential crisis” permeating developer and AI-adjacent communities. With hype swirling around new agentic AI and tools (especially “Claude Code”) and doomsayers prematurely proclaiming the end of SaaS, the Sharkey brothers offer their “proudly average” but lived-in perspective. The episode is a deep-dive—equal parts skeptical and excited—into the real capabilities (and limitations) of modern agentic AI, the rising “everything app” paradigm, whether SaaS is doomed, and the future of work with AI. All with the brothers’ characteristic banter, demystification, and no shortage of hot takes.

Key Discussion Points & Insights

1. The 2026 Existential Crisis & AI Hype (00:10–07:00)

• Michael describes two polarized “camps” in AI commentary: “the hype boys” (“agents are finally here; Claude Code can do my washing”) vs. doomsayers (“2026 existential crisis”—jobs and SaaS are dead).
• Both feel the social media-driven hysteria (“tweet boys,” “bot armies”) can lead to developer apathy or FOMO.

  “It just leads to this sense of apathy. It’s like, well, what’s the point? ... So why should I even try?” —Chris (01:17)

• When actually “doing” the practical work, much of the hype dissolves; real-world usability lags behind Twitter threads.

2. Experimenting with Agentic AI—Reality vs. Hype (02:20–05:57)

• Tools like Claude Code feel a step forward, able to work on local files and build real artifacts, but in complex use cases, the pain often outweighs the magic.
• Michael reflects that agentic models’ core abilities aren’t “radically” different than a year ago; what’s changed is packaging and accessibility.

  “Any significantly hard challenge in a big code base, I find them incredibly painful to deal with.” —Michael (03:45)

• Chris uses the analogy of “Looney Tunes” houses—impressive but structurally unsound AI-generated projects that break under real business needs.

3. Social Media, Doomscrolling & FOMO (05:57–08:02)

• Michael laments the “momentum shift” toward inflated examples with little practical substance.
• The sense of frantic progress is often an illusion; step away, come back, and “nothing’s really changed.”
• Chris feels models have hit a lull in quality (“the models are the worst they’ve been in a long time” —07:03); the real shift is more users discovering what early adopters already knew.

4. The Evolution of the AI Workspace: The “Everything App” (08:02–11:46)

• Michael details his workflow in Sim Theory, integrating email, calendars, and files into a single AI workspace.
• Predicts the AI workspace will become the “start and end point” for digital work, outshining single-function SaaS.

  “These AI apps start to become the everything app in a lot of ways, like the start and end point of the universe for workers.” —Michael (10:04)

5. Agentic Workflows: Coworking, Collaboration vs. Delegation (11:46–19:32)

• Agentic “sub-agents” (dedicated skills) can automate multi-step tasks, but interactive, collaborative workflows remain critical.
• Chris likens the process to managing human employees—sometimes you want relentless, unfettered execution (“agentic”), other times close, iterative collaboration is essential.

  “Do a bit, then stop... Like whenever I hired someone new... work on this for three hours and then stop. And I’ll check in where you’re at just to see we’re on the same page.” —Chris (15:47)

• Despite agentic automation’s lure, daily drivers remain collaborative (interactive) work; full delegation is rare and often impractical.

6. Daily Usage, Cost, and Human-in-the-Loop (19:32–24:17)

• The brothers question if agentic loops will be “everyday work” or remain occasional, high-value uses due to cost and complexity.
• Chris notes the cost problem: even at $200/month for Claude’s “unlimited” plan, it’s not scalable for large orgs.

  “There’s no way you can have people sitting around running these agentic loops all day and do it for $200 a month.” —Chris (21:31)

• Michael: productivity gains are real, but human oversight and active involvement are still essential, especially as delegation scales.
• The “replace-all-humans” narrative remains a fantasy until sustained real-world cost saves and output rival human effort.

7. The Case for Cheaper, Open Models and Multimodal Approaches (24:17–29:09)

• Chris advocates for leveraging cheaper, “lesser” models; for many workflows, advanced model knowledge is overkill.

  “We’re reaching the level where it is realistic to do this on the lesser models because the model’s own knowledge... is becoming less important.”

• The push will be toward “optimizing” these models for cost-effective widespread deployment, especially for agents.

8. SaaS Isn’t Dead (Yet)—But Is Vulnerable (29:09–38:51)

• Michael sees the AI workspace as the biggest existential threat to standalone “business SaaS,” but disruption will take years.
• He rebuts the idea that enterprises will “agentically” clone all their SaaS—maintenance overhead is real, as is the challenge of building mission-critical apps from scratch:

  “The problem is how many pieces of software am I then maintaining?” —Chris (31:37)

• Core SaaS like CRM, productivity tools, and workplace apps could be replaced piecemeal, especially as AI workspaces bake in multi-role, multi-permission enterprise-ready features.
• But: “Have you actually tried to do this?”—rebuilding and maintaining mature SaaS is much harder than it sounds.

9. The “Everything App,” App Stores & Platformization (38:51–43:39)

• Michael predicts chat-based AI workspaces will absorb more business functionality—eventually replacing emails, calendars, and CRM within the AI UI.
• SaaS vendors that don’t open up (e.g., restrictive API/“MCP” policies from Atlassian, Trello) risk being cloned and replaced by open, AI-first alternatives.

  “If a lot of these companies don’t get their act together... people are going to replace it.” —Chris (54:07)

• The rise of “app stores” for AI (add-ons, plugins, skills/MCPs) will disrupt SaaS as users bolt on what they need instead of subscribing to many siloed services.

  “To me calling it [SaaS] dead is like, quite frankly...” —Michael (58:02)

10. Proliferation of Dedicated MCPs/Skills & Proprietary Data (58:02–62:04)

• Huge opportunities exist for paid, reliable skills/MCPs (e.g., financial datasets, industry contacts).
• Existing app store attempts have fallen short, but demand for plug-and-play, reliable connectors/data feeds is strong.
• Michael and Chris both see a gold rush coming for proprietary, domain-specific MCPs.

11. Geoffrey Hinton Segment—Meme of the Week (62:10–64:35)

• Michael shares a snippet from a recent Hinton talk where Hinton, asked to stay for audience questions, deadpans: “I’d rather get back to writing my book.” (63:26)
• Chris jokes about Hinton, “the AI love rat,” and his newfound relevance (and touring Tasmania).

Notable Quotes & Memorable Moments

• “I think what it is, is more people being able to discover what other people had already been doing with the models, just via better software accessibility.” —Chris (07:23)
• “All agentic mode is: how relentless do you want it to be?... The difference is, at the end, do a quality assurance check and make sure it got what you wanted.” —Chris (15:47)
• “The other trend... is this sort of feeling... is SaaS dead?... Have you actually tried to do this?” —Michael (31:37)
• “You alone have done that four or five times [internal software projects]... Do you know how hard it is to get a working system that works in production at scale with thousands of users? It is not easy.” —Chris (32:52)
• “If you use an existing AI platform, you’ve already got all that [roles/permissions, security]—at least on Sim Theory you do.” —Chris (55:30)
• On AI agents changing lots of code: “All done by AI. I’ve changed 4,000 lines of code. Now painstakingly go through and work out which one broke my other thing... The reality is, no matter how good the context window is...” —Chris (27:45)
• “If I was running an enterprise right now, I’d be like, turn my own team into consultants: here’s all the tools, go nuts.” —Michael (49:07)
• “If we want to get the most out of it for [an] organization, we need formalized enterprise training... You’re going to remain a group of people alienated from what is possible with the technology.” —Chris (50:17)
• On Geoffrey Hinton:

  HOST: “Are you happy to stay around for a little while afterwards for any burning questions people might have?”
  HINTON: “Actually, I’d rather get back to writing my book.” (63:26)

Timestamps for Major Segments

• Opening, Existential Crisis vs. AI Hype: 00:10–03:00
• Agentic AI: Reality Check & Examples: 02:20–08:00
• AI Workspaces & Everything App: 08:00–11:46
• Agentic vs. Collaborative Workflows: 11:46–19:30
• AI in Daily Work: Limits, Loops & Cost: 19:30–24:17
• Model Costs, Cheaper Models, Multimodal: 24:17–29:09
• SaaS Threats, AI Workspace as Platform: 29:09–38:51
• App Stores, MCPs, & Disruption: 38:51–43:39
• AI App Stores, SDKs, Proprietary MCPs: 55:41–62:04
• Hinton Anecdote (Comic Relief): 62:10–64:35

Episode Takeaways

• The practical impact of agentic AI is substantial but falls short of the hype—workflows are still maturing.
• The trend is toward powerful, modular AI workspaces becoming “everything apps,” absorbing SaaS functionality over time.
• The “SaaS is dead” narrative is overblown, but disruption is inevitable for slow-moving incumbents.
• Widespread effective adoption hinges on cost, model flexibility, proper interfaces, and (most importantly) training—not just tools.
• Open platforms and plug-and-play “skills”/MCPs—especially for proprietary data—are a huge, currently underexploited business frontier.
• The future is less about AGI replacing humans, and more about empowering people to work in collaboration with AI.

Final Thoughts

Both hosts agree: AI’s trajectory is more confusing, but also more hopeful, than ever. Organizations must become intentional and strategic about integrating AI—focusing on real-world, sustainable use, and user enablement, not hype. SaaS is on borrowed time in many areas, but the end isn’t nigh—yet.

“Everybody’s still learning so I think everyone needs to look at like what are you trying to get out of it and how can you empower the most people to be using it?... I don't want to come across like I'm doubting anything. I think it’s unbelievable stuff. I’m so excited to be back and working on everything. It’s just that I really feel like it's the most confused time about what, what your daily workflow looks [like].” —Chris (66:39)