Episode

2025 was the year of agents, what's coming in 2026?

Podcast: Practical AI
Published: Jan 9, 2026
Duration seconds: 3075
Processing state: processed
Canonical source: https://share.transistor.fm/s/e7fda8ce
Audio: https://pscrb.fm/rss/p/dts.podtrac.com/redirect.mp3/media.transistor.fm/e7fda8ce/c218cdfa.mp3
JSON: /v1/public/podcasts/practical-ai/episodes/2025-was-the-year-of-agents-what-s-coming-in-2026
Markdown: /podcast/practical-ai/2025-was-the-year-of-agents-what-s-coming-in-2026.md

Actions

POST https://stenobird.com/v1/public/podcasts/practical-ai/episodes/2025-was-the-year-of-agents-what-s-coming-in-2026/transcription-requests
Idempotently request low-priority transcript generation for this episode.
GET https://stenobird.com/podcast/practical-ai/2025-was-the-year-of-agents-what-s-coming-in-2026.md
Read the agent-friendly Markdown representation of this episode resource.

Summary

Reflecting on the transition from LLMs to autonomous agents in 2025, this episode forecasts the shift toward orchestration and energy-constrained computing in 2026. The hosts analyze how reasoning models and regulatory compliance will define the next frontier of AI implementation.

Topics

AI Agents
Reasoning Models
AI Orchestration
Machine Learning Infrastructure
AI Compliance
NIST Standards
Energy Constraints
Autonomous Workflows

Highlights

Main idea: 2025 marked the transition from simple model assistance to autonomous agentic workflows
Failure mode: Extended reasoning tokens improve accuracy but introduce significant latency and increased inference costs
Practical takeaway: The next major bottleneck for AI scaling is shifting from GPU availability to power and energy constraints
Main idea: Future competitive advantage lies in orchestration and managing complex, multi-service AI workflows
Practical takeaway: Achieving regulatory compliance (like NIST 600-1) is becoming a primary driver for enterprise AI adoption

Chapters

1:00 The Shift to Agents: A look back at 2025 as the year AI transitioned from basic models to autonomous agents.
4:45 Finding Value in the Hype: Discussing the difficulty of finding successful, practical use cases amidst the agentic hype.
8:30 The Impact of Reasoning Models: Analyzing how reasoning tokens and hybrid models are reshaping workflows and introducing latency.
20:00 Multimodal Realities: Evaluating the current state of multimodal inputs versus the continued dominance of text-based outputs.
27:20 The Energy Bottleneck: Why power availability and energy consumption are becoming more critical than GPU counts.
39:20 The Future of Orchestration: Predicting that the next year will focus on application-specific agentic systems and complex orchestration.
47:20 Compliance and Complexity: The challenge of building compliant AI systems in regulated industries and the opportunity for consolidated solutions.