The New AI Superpowers: Focus and Followthrough

Last updated: July 27, 2026

The conventional pitch for AI coding tools is seductive: finish your work 10x faster, reclaim your evenings, and finally build all those side projects you have been collecting. Rick Manelius, a repeat startup founder, tried exactly that. He queued up Claude in 5-minute increments between calls, let it rip, and started shipping project after project. The result was not liberation. It was 40 proof-of-concept projects, a growing sense of burnout, and the uneasy realization that AI had turned him into a machine that converts time into unfinished work.

His essay The New AI Superpowers: Focus and Followthrough landed on the Hacker News front page on July 26 and collected 168 points and 49 comments in its first day. The argument is simple and uncomfortable: AI productivity gains are real, but they are horizontal by default. Every new project you spin up is a new open loop, a new responsibility, a new thing to tend. The tool that was supposed to free you instead fills every available slot in your attention.

What HN Is Saying#

The HN thread on this piece is worth reading because it mostly agrees with the diagnosis but pushes hard on the causes and solutions. A few themes dominated.

The AI-as-config-bandage pattern. The top comment thread, from LogicFailsMe, describes using AI not to build products but to fix the supporting stack: "config, container, and installation issues so I can spend my time coding instead of struggling to fix mostly other people's mistakes." This is a recurring pattern across the discussion - developers are using AI to absorb the ambient complexity of modern tooling (Python wheels, containers, virtual environments, fragmented APIs) rather than to multiply output. One commenter observed that it might actually be a mistake to have let software complexity grow this much in the first place, and that AI is a bandage on a wound the industry inflicted on itself.

Skepticism about 100x claims. Several commenters pushed back on the premise that AI delivers 100x productivity. "I will never believe the premise of 100x boost from AI," one wrote, noting that the code is "nowhere fit for production" and that the last 10% needed to ship reveals the first 90% as "hacks on top of hacks." Another pointed out that having a backlog of 99% projects instead of 0% projects is not really an improvement if nothing ever ships. This connects directly to the article's central insight about the last 1% - the difference between a partial eclipse and a total eclipse, as Garry Tan put it.

Essentialism rediscovered. Multiple commenters observed that the article is essentially a rediscovery of creative disciplines that writers, designers, and artists have practiced for decades. One noted that in writing, the advice is to "kill your darlings" - do not let a paragraph you love derail the whole story. Another pointed out that the same problem has always existed in code: developers treat every generated feature as precious, accumulating clutter that leads to "Implementation Fatigue." The framework of Essentialism, which Manelius cites from Greg McKeown, resonated with readers who recognized the pattern from their own experience.

The organizational value function is lagging. One of the most insightful comments argued that AI has created a dynamic where everyone believes every problem is "a couple hours" with AI, leading to teams where everyone builds incompatible versions of the same beginner-level software. The commenter noted that if a solution has zero external dependencies, it is probably a toy. The organizational incentive structure has not caught up with the technology - "proof of concept" used to correlate with "proof of work," but now the correlation is gone.

Counterpoint: some people love having many projects. Not everyone agrees with the burnout thesis. One commenter described launching one side project per week for months and finding it energizing, not draining. "The nagging feeling that I am missing out, did not try, should be working on something, seeing competitors launch my idea - all gone." Another described switching to a fixed cycle of writing specs, launching background agents, reviewing, merging, and releasing - and seeing feature velocity go up without the stress.

Why This Matters for Developers#

Manelius's essay touches a nerve because it describes a problem that is structurally new but emotionally familiar. Developers who have been in the industry long enough recognize the feeling of productivity rising while satisfaction falls. The difference is that AI accelerates both sides of the equation at once.

The real insight is not that AI causes burnout. It is that AI removes the natural friction that used to force prioritization. Before AI, starting a project required a significant time investment - you had to really want it. Now, starting is nearly free. The result is a flood of half-built things, each demanding attention, each an open loop in your cognitive stack.

This is the same observation at the heart of the human-in-the-loop is tired argument from Pydantic. When AI eliminates the cognitive rewards of coding - the satisfaction of solving a problem, the feeling of mastery, the moment something clicks - what is left is the grind of supervision without the joy of creation. Burnout in the age of AI is not about working too many hours. It is about working on too many things without the feedback loops that make work meaningful.

The practical takeaway is that focus is a skill you need to practice deliberately. The AI-native development workflow that separates developers getting 5-10x gains from those getting 20-30% is not about using more tools. It is about using fewer, better. The same principle applies to projects. The developers who will win in this era are not the ones who ship the most features. They are the ones who ship the right features, finished.

The over-editing problem is another facet of the same pattern. When agents generate massive diffs for tiny fixes, they create the illusion of progress while producing mostly noise. The discipline of focused, minimal changes is more valuable than ever - not because AI cannot handle large changes, but because the human cost of reviewing and integrating those changes is still real.

The concept of good tools being invisible maps directly here. When a tool demands attention - when it generates output you have to triage, review, and clean - it is not actually making you more productive. It is just transferring cognitive load from creation to curation. The best AI setups are the ones that disappear into the workflow, not the ones that generate the most impressive diffs.

And the overnight agent workflow offers a structural solution to the attention problem. By scoping agent tasks to a single focused session with a clear deliverable, you avoid the sprawl of 40 concurrent proof-of-concept projects. The constraint is the feature, not the tool.

Sources#

• Manelius, Rick. "The New AI Superpowers: Focus and Followthrough." July 26, 2026. https://www.rickmanelius.com/p/the-new-ai-superpowers-focus-and
• Hacker News discussion. July 26, 2026. https://news.ycombinator.com/item?id=49057877
• McKeown, Greg. Essentialism: The Disciplined Pursuit of Less. Crown Business, 2014.
• Tan, Garry. "Partial vs Total Eclipse." https://x.com/garrytan/status/2062760454649487491
• Summers, Laura. "The Human-in-the-Loop Is Tired." Pydantic. Developers Digest analysis

Continue Reading#

• The Human-in-the-Loop Is Tired: Pydantic on AI Dev Burnout - Laura Summers argues that LLM-assisted programming eliminates the cognitive rewards that made coding satisfying, leading to structural burnout.
• The AI-Native Development Workflow - The five-layer stack that separates developers achieving 5-10x gains from those stuck at 20-30%.
• Good Tools Are Invisible - Ginger Bill on essentialism in engineering and the trap of mistaking busy for productive.
• Over-Editing: Why Your AI Coding Agent Rewrites What Isn't Broken - A quantified look at the quality gap between agent output and production-ready code.
• Ship Code While You Sleep: The Overnight Agent Workflow - How to scope agent tasks for focused, verifiable results without accumulating half-built projects.
• Microsoft's CLI Coding Agent Study: Adoption Is a Workflow Problem