The twilight of the chatbots

2026-02-18 · Source: One Useful Thing · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, short

Summary

AI model capabilities and release rates are accelerating, with frontier models like Claude Fable and GPT-5.6 facing government access restrictions despite rapid advancements. Assessments from METR, the UK's AI Security Institute, and GDPval indicate "better than exponential" growth in AI's ability to perform human work. For example, Opus 4.7 autonomously built software in 14 hours for \$251, a task estimated to require 2-17 weeks of human engineering. The author's experiments with Fable showed 9 hours of autonomous work on complex software projects. While US companies like Anthropic, OpenAI, and Google lead with frontier models, Chinese open-weights models, though lagging 6-12 months, also show exponential improvement and are cheaper. This evolution is shifting AI usage from co-intelligence with chatbots to managing autonomous agents, which leverage specialized harnesses and apps. An OpenAI study revealed widespread agent adoption across coding, legal, and HR, with a quarter of its workforce running at least four agents weekly. Domain expertise, not profession, correlates with greater agent success. This exponential growth creates "shocks" in policy and markets.

Key takeaway

For Directors of AI/ML evaluating strategic roadmaps, recognize that AI capabilities are evolving exponentially, rendering plans from early 2025 or prior obsolete. Your teams should pivot from chatbot-centric co-intelligence models to deploying and managing autonomous AI agents for complex, long-running tasks. Prioritize developing deep domain expertise within your staff, as this is the primary driver for maximizing agent effectiveness and achieving significant productivity gains. Embrace a management-focused approach to AI integration.

Key insights

AI capabilities are accelerating exponentially, driving a shift from interactive chatbots to autonomous, expert-managed agents.

Principles

• AI capability growth follows an exponential curve.
• Domain expertise enhances AI agent effectiveness.
• Open-weights models offer cost-effective, rapid improvement.

In practice

• Assess AI agents for complex, multi-step autonomous workflows.
• Prioritize domain-specific training for AI system operators.
• Explore open-weights models for budget-conscious AI deployments.

Topics

• AI Agents
• AI Capability Growth
• Open-Weights Models
• AI Adoption Strategy
• Domain Expertise
• AI Evaluation Benchmarks

Best for: Investor, CTO, AI Architect, Director of AI/ML, VP of Engineering/Data, Consultant

Related on AIssential

• MLWhiz Weekly AI/ML Newsletter # 2 — The AI industry has shifted from a model capability race to a platform war for agentic workflow ownership.
• You Are the Pilot Now — AI agents autonomously execute complex, multi-step tasks by planning, using tools, and adapting, unlike simple chatbots.
• not much happened today — The AI landscape is rapidly evolving towards specialized, agentic systems, with open-weight models gaining significant ground against proprietary solutions.
• #211: GPT-5.5, ChatGPT Workspace Agents, The Messy Reality of Agents & Google Cloud Next — AI is rapidly shifting from chat-based models to autonomous, task-oriented agents, with major players introducing new platforms.
• This Week in AI — AI advancements are rapidly commoditizing capabilities while enabling new forms of automation and creative expression.
• How We Use AI Is Changing — AI usage is evolving from basic chat to agentic systems and "loops," creating a significant performance gap between casual and power users.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by One Useful Thing.