A Brain is Alive in a Computer

2026-03-10 · Source: There's An AI For That · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Robotics & Autonomous Systems · Depth: Intermediate, extended

Summary

This intelligence brief critiques an essay by AI startup entrepreneur Matt Schumer, which claims AI progress is rapidly accelerating, particularly in code generation. The analysis, provided by Cal Newport, refutes Schumer's assertion that AI models experienced exponential growth in 2025, stating that general capability improvements slowed post-GPT-4, leading companies to focus on narrow, post-training tasks. Newport highlights that while coding agents like GPT-53 Codex and Opus 4.6 show incremental progress in specific programming tasks, they do not enable programmers to fully automate app development by simply describing requirements and returning hours later. He also dismisses the idea that AI writing code for itself will lead to recursive self-improvement and a general AI takeover, labeling it "gradeA nonsense" and a misinterpretation of current AI capabilities.

Key takeaway

For CTOs and VPs of Engineering evaluating AI integration, recognize that current AI tools, while useful for specific programming tasks, do not yet support full automation of app development. Your teams should plan for heavily supervised AI use, focusing on clear specifications and extensive testing for AI-generated code, rather than expecting a "set it and forget it" workflow. Avoid alarmist narratives about exponential AI growth or self-improving agents, as these claims are not supported by current technological realities.

Key insights

Current AI progress is incremental and specialized, not a general exponential acceleration or self-improving loop.

Principles

• General AI capability growth has plateaued.
• AI progress is often task-specific and incremental.
• AI models do not self-improve or invent new intelligence models.

Method

AI companies shifted from general pre-training scaling to post-training and fine-tuning for specific applications, such as coding agents, to achieve incremental improvements and find viable markets.

In practice

• Use AI for tedious, structured coding tasks.
• Extensively test AI-generated code for accuracy.
• Focus AI applications on narrow, well-defined problems.

Topics

• AI Investment
• AI Progress Critique
• AI Code Generation
• Computational Neuroscience
• AI Office Productivity

Code references

• sipeed/picoclaw
• NousResearch/hermes-agent
• msitarzewski/agency-agents

Best for: CTO, VP of Engineering/Data, Director of AI/ML, General Interest, Software Engineer, AI Engineer

Related on AIssential

• The AI Wake-Up Call Everyone Needs Right Now! — AI's exponential progress, driven by self-improvement, is rapidly transforming professional work and societal structures.
• How to build proactive agents & self-improving company (Fully explained) — AI native loops enable self-improving companies by integrating continuous feedback for autonomous learning and operational enhancement.
• The Time Savings Era of AI is Over — AI benefits are shifting from time savings to increased output and new capabilities, driven by agentic and multimodel usage.
• not much happened today — AI agents are expanding beyond APIs to direct computer control, while model training and evaluation become more efficient and industrialized.
• I Wake Up to ‘AI Is Replacing Everyone’ Every Day. But Then I Read AI Builds AI — AI models are now contributing to their own development, accelerating autonomous evolution.
• OpenAI's chief scientist says AI progress has been "surprisingly slow" and promises big leaps ahead — OpenAI expects rapid AI advancements, driven by models like GPT-5.5, despite current progress being "surprisingly slow."

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by There's An AI For That.