Podcast Summary: Practical AI

Episode: 2025 was the Year of Agents, What's Coming in 2026?

Host: Daniel Whitenack (B), Chris Benson (C)
Date: January 9, 2026

Episode Overview

This episode kicks off 2026 with hosts Daniel Whitenack and Chris Benson reflecting on the evolution of artificial intelligence over the past year—particularly the rise of AI agents—and discussing the practical realities, challenges, and predictions for the coming year. The conversation centers around:

• The transition from models to assistants to agents in AI
• The realities and hype behind enterprise adoption of AI agents
• The transformative (and limiting) effects of AI on development workflows
• Advances in multimodal and reasoning models
• Infrastructure, power, and geopolitical impacts of expanding AI
• The ongoing role of predictive models vs. generative AI
• Predictions for 2026, including the increasing complexity and fragmentation of AI systems, and the advent of AI-maker culture

The tone is informal, thoughtful, and balanced between optimism and caution.

Key Discussion Points & Insights

1. 2025: The Year of AI Agents

[02:00-14:07]

• Shift in Focus: 2025 marked a distinct transition in AI discourse:
  “For a while we talked about models, then we kind of talked about assistants, and then we really kind of transitioned to talking about agents. Agents are autonomous AI. That was a key theme of 2025.” (B, [02:14])
• Hype vs. Reality: There was significant hype, but adoption numbers were exaggerated:

  “I've seen some crazy things like, like, from, like, nobody, you know, successfully using them all the way to, like, 70% of all existing organizations are now using AI agents, which I'm like, totally, like, BS, you know... It's a long way from the truth.” (C, [04:01])

• Success Factors: Effective use of AI agents depends on finding the right use case, having domain expertise, and understanding integration.

  “Looking into the right use case and getting the right people to address it and having a good business case for it makes a lot of difference.” (C, [05:14])

Notable Moment:

• Anecdote on Developer Impact:
  Chris shares how recent model advances now allow coding agents to perform “senior level coding really well without mistakes, …it's kicking butt in Rust now. …My workflow has changed dramatically in the last two months in terms of understanding how to effectively use coding agents to do that.” (C, [07:01])

2. The "Alien Tool" Effect: Personal and Organizational Transformation

[08:18-15:51]

• AI as a Force Multiplier:
  Daniel recounts investors noting massive product output increases without team expansion:

  “Product wise, you all were able to advance so much more quickly without expanding your team… The pace of development… is significant. It's significant enough to be noticed...” (B, [09:03])

• Chris’ Aha Moment:

  “In the matter of six minutes it produced what I would have at least six weeks of work, at least six weeks of work in just a matter of a handful of minutes… I had a large project laid out in VS code… I've never had that big of an aha moment in coding.” (C, [11:07])

• Not Plug-and-Play:
  The technical and domain expertise required for effective agentic AI is substantial:

  “There had been literally many hundreds of prompts before that… At no point was the AI leaving me behind, it would open up new doors, but I had to walk through those doors, take the learnings and develop the next prompt from it.” (C, [14:07])

3. Multimodal and Reasoning Models: Progress, Limitations, and Annoyances

[17:51-25:09]

• Multimodal Adoption Still Input-Heavy:

  “In terms of multimodal on the input side, I still very much do not interact with people that are doing kind of multimodal on the output side. …Not really multimodal output. Maybe the exception to that might be synthesized speech…” (B, [18:54])

• Rising Importance of Reasoning:
  The so-called "reasoning era" is dawning, but the models don’t actually “reason”—they just mimic it via “chain-of-thought” text.

  “Reasoning models are mislabeled because they don't reason about anything, they just produce text… What is interesting is that they produce a segment of text that imitates or mimics reasoning…” (B, [22:22])

• Latency Trade-offs:

  “Often, like in real business applications… you really don't want those reasoning tokens because they take so dang long. Right? There's so much latency… It's kind of annoying that a lot of these later models have these.” (B, [23:38])

4. Infrastructure, Power, and Geopolitical Dynamics

[25:09-32:08]

• Shift from GPU Scarcity to Power Scarcity:

  “For years we talked about the limitation of having enough GPUs… and now it is power… every prompt that I choose to make in a quote unquote reasoning fashion is going to be much more expensive in terms of power consumption." (C, [25:40])

• Anecdotes Illustrate Scale of Shifts:
  Daniel discusses speculators buying power plants out of belief decommissioned ones will be needed again, and the backlash against large chip assembly plants due to local infrastructure strain.

  “You see that dynamic here, right? In China… they're just going to put a power plant there, right? …That’s how this has then filtered into this geopolitical space… power and AI and chip manufacture and onshoring—all of this is what's driving now the political conversations.” (B, [27:32])

• Policy is Now AI-Driven:

  “We've moved from … talking about … how might governments regulate AI... to now … AI is the topic that is driving some of those policy decisions…” (B, [30:23])

5. Predictive Modeling: The Understated Workhorse

[32:08-38:16]

• Generative (Gen AI) Plateau, Predictive Keeps Advancing:

  “Gen AI models have sort of plateaued on this transformer architecture… but predictive models still continue to advance… across industry these models still continue to provide amazing ROI and get better and better.” (B, [32:38])

• AutoML to Augmented ML:

  “…There is actually this realization of maybe a better automl or ... augmented analytics or augmented ML … highly capable tools... can actually be tied in as tools into a generative AI model that orchestrates amongst all of those and reasons over how to use those.” (B, [33:28])

Notable Moment:

• Yann Lecun Leaves Meta; Transformers’ Ceiling Discussed

  “One of the three godfathers of AI… Yann Lecun, has left Meta… And one of the things that he has talked about… is the fact that Transformers had a limited ceiling…” (C, [36:18])

6. Skills, Architecting Agentic Systems, and the "AI Engineer"

[38:16-43:09]

• The Next In-Demand Role:

  “My prediction… is that those practitioners that have the capability… to actually architect that agentic system regardless of model… I think that is a wildly powerful combination.” (B, [38:21]) “You're the human at the center of a great symphony of AI agents and you have to learn to conduct those agents in that symphony to produce way more than you could have ever done last year.” (C, [40:46])

• Recommendation:
  Data scientists, software developers, and others should seek skills in tool integration, orchestration layers, and domain-specific system building.

7. Predictions and Looking Ahead to 2026

[43:09-49:49]

• Chris’ Prediction: The AI Maker Era

  “As you have these orchestrators with the tooling around them with predictive models now kind of enabled through agentic systems… the maker world is really starting to see that as a possibility… So my prediction is we see the very beginning of the AI maker era come about at a consumer level.” (C, [43:17])

• Daniel’s Predictions:
  • Model Commoditization & Complexity:

    “Models have been quite commoditized. The increases in performance… have plateaued. Open source models have essentially caught up… The most relevant thing is… flexibility, not getting lock in, the ability for you to use a bunch of different models, the ability for you to, you know, construct a system.” (B, [45:59])

  • Fragmentation & Management Complexity:

    “What is problematic is… I need all these tools, I need to connect them in a certain way. That becomes increasingly complicated… expansion of complexity in these AI systems, not because the models are not capable, but because the model is actually no longer the blocking point…” (B, [47:09])

  • Winners Will Offer Simple, Compliant, Verticalized Solutions:

    “I think some of the winners in this space are going to be those that come to that complexity and tell you, hey… here's a consolidated quick time to value way for you to get X or Y, whether that be a verticalized AI solution, a secure AI solution, whatever that might be.” (B, [48:49])

Notable Quotes & Memorable Moments

• On the velocity of AI innovation:

  “Even I at moments are feeling a bit left behind with how fast this is changing.” —Paraphrasing Andrej Karpathy’s reflections (C, [06:15])

• On agentic AI as a force multiplier:

  “These agentic workflows… are transformative in ways that are legitimately transformative, multiplicative, however you want to say that.” (B, [12:12])

• On learning and prompt iteration:

  “At no point was the AI leaving me behind, it would open up new doors, but I had to walk through those doors, take the learnings and develop the next prompt from it.” (C, [14:19])

• On the energy arms race:

  “Some countries are now invading other countries and taking their oil… Power is, is the thing that people are talking about…” (C, [29:44])

• On the new AI skills imperative:

  “You're the human at the center of a great symphony of AI agents and you have to learn to conduct those agents…” (C, [40:46])

• On the coming fragmentation:

  “It’s no longer about, I'm going to get the best model. And now my company has AI and I'm set for the future. That's actually the easiest thing… expansion of complexity in these AI systems… the model is actually no longer the blocking point of the whole thing or the single thing in the system.” (B, [47:09])

Timestamps – Section Reference

• [02:00] – 2025: The Year of Agents
• [08:18] – The “Alien Tool” Impact - Personal/Business Transformation
• [17:51] – Multimodal and “Reasoning” Models, Latency Trade-offs
• [25:09] – Hardware, Infrastructure, and Power Constraints
• [32:08] – Predictive vs Generative AI, Tool-Orchestration
• [38:16] – Skills & Role Evolution: “AI Engineer”/System Integrator
• [43:09] – Chris & Daniel Predict 2026 — Maker Era, Complexity, Winners
• [49:49] – Wrap-up & Looking Forward

Final Thoughts

The episode is an engaging and grounded reflection from two industry veterans. Listeners new and old will come away understanding:

• Why AI agents, not just models, will shape 2026
• The human expertise and adaptation required for real ROI
• That the infrastructure and societal impact of AI extends far beyond technical advances
• The fragmentation and complexity coming for anyone seeking to deploy AI—meaning integrators, not model developers, will be the key technical heroes of the next year
• The rise of an “AI maker era,” with capabilities increasingly available at the consumer and hobbyist level

The hosts close with encouragement for listeners to connect, experiment, and embrace the changes heading into 2026.