Claude Code creator Boris Cherny says loops are “just as important and as big a step” as the jump from source code to agents. At Meta’s @Scale conference, the first audience question was about loops. The AI world just found its next paradigm β and it’s the one Business Engineer has been writing about all month.
Loops take agentic AI a step further. Instead of a human prompting an agent to do a task, agents prompt other agents in continuous background cycles β endlessly improving, checking, and rebuilding without human intervention.
Cherny described his own setup: one agent continually looks for ways to improve code architecture, while another hunts for duplicated abstractions that can be unified. They run in the background, always on, always improving. The human doesn’t trigger them. They trigger themselves.
The evolution in one line:
Era 1
Human writes code
Era 2
Human prompts agent β agent writes code
Era 3 β Loops (now)
Agent prompts agent β agents improve code continuously, in the background, without human trigger
The key insight: The shift from agents to loops is the shift from tool to system. An agent does what you ask. A loop does what it determines needs doing β then does it again, better. That’s not assistance. That’s autonomous improvement.
The Week That Made Loops Undeniable
Cherny’s statement doesn’t exist in isolation. This week alone:
Perplexity Brain β agent builds a context graph during the day, synthesizes overnight, loads improvements into the next session. +25% accuracy from the loop.
Sakana RSI Lab β AI systems that improve the training methods for future AI systems. LLM-Squared: models training models. Loops at the architecture level.
Cursor /automate β describe an automation in natural language, agent wires up triggers and instructions. Background agents that run without human prompting.
The Builder-PM Manifesto β “The loop replaces the roadmap.” The Builder-PM’s continuity comes from the validation loop, not from stakeholder reviews.
Four independent signals in one week, all pointing the same direction: the competitive edge is moving from the model to the loop around it.
The Structural Read
LOOPS ARE THE HARNESS THEORY ENDGAME
The Harness Theory argued the edge is the orchestration system around the model, not the model itself. Loops are what happens when the harness stops needing a human to operate it. The harness runs itself. The human frames what the harness gets pointed at. That’s the principal/operator split in code.
Prompt engineering was the 2023 skill. Agent orchestration was 2025. Loop engineering β designing the feedback cycles that agents run autonomously β is the 2026 skill. It’s already being taught. The McKinsey Skill Change Index showed execution skills automating fastest. Loops are why: the execution layer now runs itself.
THE COMPOUND ADVANTAGE IS NOW AUTOMATIC
An agent that runs once gives you leverage. A loop that runs continuously gives you compounding. Every cycle, the code gets better, the architecture gets cleaner, the abstractions get tighter. The person running loops compounds faster than the person running agents β not by working harder, but by designing better feedback systems.
The Bottom Line
Boris Cherny β the person who built Claude Code β says loops are as important as the jump from code to agents. Perplexity built a loop that improves overnight. Sakana built loops that train models on models. Cursor shipped /automate for background loops. The pattern is converging: the next era of AI isn’t about what the model can do. It’s about what the model does when you’re not watching. The loop is the harness that runs itself.
Business Engineer Framework
The Harness Trilogy β Why the Loop Is the Harness That Runs Itself
The Harness Trilogy mapped the shift from operating to framing. Loops are the final step: the harness no longer needs an operator. It needs an author β someone who decides what it gets pointed at.
Gennaro is the creator of FourWeekMBA, which reached about four million business people, comprising C-level executives, investors, analysts, product managers, and aspiring digital entrepreneurs in 2022 alone | He is also Director of Sales for a high-tech scaleup in the AI Industry | In 2012, Gennaro earned an International MBA with emphasis on Corporate Finance and Business Strategy.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
