How I use AI: Oct 2026 Perspective

The Shift to Agentic Workflows

As of October 2026, the paradigm of AI-assisted development has shifted from simple autocomplete to agentic coding. Rather than writing code line-by-line, developers now manage multiple AI agents—such as Claude Code, OpenCode, or Codex—that handle feature implementation, bug reproduction, and testing within isolated Git worktrees. These agents significantly accelerate prototyping and allow for rapid experimentation, yet they are not a replacement for human engineering. Left unguided, agents often produce over-architected, unreadable, or "vibe-coded" solutions. Human taste, architectural oversight, and rigorous manual validation remain critical to ensuring the quality and maintainability of the final output.

The Reality of Agentic Fatigue

While agentic coding offers immense speed, it introduces a new, exhausting form of cognitive load. Managing multiple simultaneous conversations with agents across different tasks can be overstimulating. Users often report a decline in the "meditative" joy of programming, as the act of crafting code is replaced by the management of automated systems. Furthermore, there is a persistent risk of "AI fog," where the effort required to review, debug, and correct agent-generated code can exceed the effort of writing it from scratch.

Research and Learning

Generative AI has become the default interface for search and literature reviews, often outperforming traditional search engines by synthesizing nuanced data. However, AI remains an imperfect tool for novel research. While models are excellent at explaining well-known concepts and acting as Socratic tutors, they can lead researchers astray when tasked with complex, original problem-solving. Users are cautioned that LLMs may exhibit subtle biases or "hallucinate" architectures, necessitating a reliance on human experts and foundational academic literature to ground research.

The State of AI Content Generation

AI-generated writing remains largely mediocre, often characterized by repetitive, vapid prose that lacks genuine insight. While tools are useful for finding specific words, they are generally ineffective as editors for high-quality work, as they tend to regress to the mean and lack critical nuance. Similarly, the landscape of AI art and video is increasingly saturated with "slop"—low-effort, commercially driven content designed to capture ad revenue. Conversely, AI music has found a niche as a lens for human creativity, though the community remains wary of commercialized, repetitive synthetic content.

Community Insights and Counterpoints

Discussions surrounding modern AI usage highlight significant skepticism regarding the claim that "coding is solved." Critics argue that:

"There is no substitute for reading the code in many cases. I find incorrect assumptions baked into the code (the ultimate spec) and missed by the agents and tests daily."

Others point out the economic and practical trade-offs of current workflows:

"$3,000/month is nuts! For $36k (1 yr of spend) you can buy a system with dual rtx pro 5000... what you get for these systems is way more than $3k/mo of tokens, because you can run it 24/7/365."

There is also a broader philosophical debate regarding the value of human skill. Commenters noted a double standard in how society views AI-assisted engineering versus AI-assisted art, questioning why we celebrate the democratization of software creation while mourning the loss of traditional craftsmanship in music and visual arts.