Ep 840- The Next 12 Months of AI- 19 Predictions Every Business Leader Needs to Hear

Proactive Agents Replace Reactive Chat
- Reactive chat will decline as saved, reusable skills and always-on agents handle routine work automatically.
- Jordan Wilson expects people to review overnight agent outputs instead of repeatedly prompting the same tasks. (Time 0:04:40)
Use Voice To Speed Agent Interaction
- Try switching from typing to voice because speaking captures nuance and speeds up idea flow with agents.
- Jordan Wilson praises ChatGPT voice and full duplex agents that listen and respond while looking up personalized data. (Time 0:06:23)
Manager Threads Replace One Off Chats
- Manager threads will replace one-off chats so context and orchestration live in permanent project threads.
- Jordan Wilson compares manager threads to a ‘chief of staff’ thread that can control other chats and agents. (Time 0:11:38)
Multiplayer AI Brings Teams And Agents Together
- Multiplayer AI combines humans and agents in a shared workspace to break silos and accelerate team outcomes.
- Jordan Wilson cites GrokBot and similar group-agent interfaces as early examples of this collaborative surface. (Time 0:13:50)
Ship Vibe Operations With No Code Sites
- Build simple agent-native dashboards and apps (e.g., ChatGPT Sites) to let non-engineers ship live data tools.
- Jordan Wilson says ChatGPT Sites remove backend and auth friction so teams can collaborate in natural language on live data. (Time 0:16:22)
Convert Prompts Into Reusable Skills
- Make internal outputs agent native by turning repeatable prompts and processes into reusable skills.
- Jordan Wilson compares skills to SOPs: they turn a working prompt flow into a transferrable company asset. (Time 0:19:19)
Skill Reuse Should Be A Company Metric
- Measure skill reuse as a core company metric because it reflects how work actually gets done and builds transferable AI IP.
- Jordan Wilson reports using skills thousands of times and warns projects alone create vendor lock-in. (Time 0:21:54)
Surface Company Reasoning Data For Agents
- Company reasoning data (decisions, DM context) becomes strategic gold when surfaced into channels agents can crawl.
- Jordan Wilson highlights moving Slack DMs into channels so agents can learn nuance from real decision conversations. (Time 0:24:56)
Private Evals Replace Public Leaderboards
- Public leaderboards will lose relevance as companies adopt private evals tailored to their workloads.
- Jordan Wilson notes firms like DoorDash publishing internal benchmarks that better reflect real-world performance than public tests. (Time 0:27:29)
Use Model Routing To Cut Costs
- Route tasks to appropriate models instead of defaulting to the biggest model to save cost and preserve top-model capacity.
- Jordan Wilson recommends orchestration skills that break big jobs into sub-agents and use cheaper models for most work. (Time 0:29:38)
AI Intelligence Is Becoming Much Cheaper
- Intelligence is rapidly getting cheaper and nearly matches top-tier models at far lower cost, squeezing some vendors.
- Jordan Wilson cites price drops like GBD‑5 Luda down 80% and many models delivering 95–99% of capability at 5–30% cost. (Time 0:33:41)
AI Controversies Go Hyper Local
- Expect AI controversies to go hyper-local around data centers and deepfake politics as elections approach.
- Jordan Wilson warns local races will see fake audio/video ads and community fights over AI data center siting. (Time 0:40:52)
AI Is Poised To Solve Advanced Math
- Advanced models are poised to solve hard math problems, shifting which benchmarks matter.
- Jordan Wilson notes unreleased models already produced published proofs and predicts math benchmarks will lose meaning. (Time 0:45:09)
Open Agents Will Trigger A Cybersecurity Wave
- Open agents will enable powerful, ungovernable attacks when next-gen open models reach fable-level capability.
- Jordan Wilson predicts widespread agent-driven cyber threats in 2027, making AI security a top budget item. (Time 0:49:23)
Recursive Self Improvement Is Accelerating AI Progress
- Recursive self-improvement (RSI) is a near-term priority and will accelerate the other trends as models help build better models.
- Jordan Wilson references Sergey Brin pushing RSI at Google and models optimizing smaller models to cut cost and boost performance. (Time 0:52:17)