AivexaNewsSearch
AI news for builders and product teamsChecked every hour

AI is overwhelming and I’m doing it wrong

Collected Oct 29, 2026

The writer behind the Ben's Bites newsletter published a candid post arguing that the current pace of AI product and model releases has become noise rather than signal, and that he has been personally caught up in it. He notes that when he started the newsletter it went out five days a week, and he cut it to two days, accepting a revenue hit, because the volume was so overwhelming. He says the noise has not gone away and may in fact be louder now. He points specifically at people running twenty personal agents to manage social content or daily life as a symptom of the moment.

The post draws on two pieces he read the previous day, one titled The Sane Asylum by Colossus and one titled The Difference Between "Can" and "Should" by Jack Raines. The Colossus excerpt mocks the relentless pressure of the current work culture, describing people grinding through long hours, skipping rest weeks, comparing how much their agents got done, chasing virality on X, angel investing and buying property in the Bay Area, with ambition described as Icarus-like and capacity redlining. The Raines excerpt makes the point that firing up a coding agent such as Claude Code, Codex or GrokBot, or dictating a request through Wispr Flow to build a daily newsletter summarizer, has made it trivially easy to start things — which sharpens the gap between what is possible and what is worth doing.

The author turns that lens on his own habits. He lists projects he has started and left unfinished: a course roughly 70 percent complete, a tool for non-technical users built on personal agents at about 80 percent, two outstanding funds with one still needing fundraising, and a new sponsor model he has yet to pitch to companies. He describes his own motivation as driven by play rather than business logic, and argues that play is what is missing for many people right now, with companies and peers pushing for more output without a clear reason.

His critique of the agent wave is framed as a hiring analogy: a company should hire when there is a real pain that hiring solves, so it is odd to be hiring and then managing multiple agents for problems that were never that painful. He also invokes the widely cited idea that people can hold only about four things in their heads at once, while conceding it may be a stretch in his case. He contrasts that with top engineers who can churn out large volumes of work and multitask, and says those are the kinds of companies he wants to invest in, since that may be what it takes to build a generational trillion-dollar company. For people who want to move fast but remain considered, he suggests there has to be another way.

He pairs that argument with a concrete editorial decision: he recently unsubscribed more than 60,000 people from the newsletter because they were not opening it, even while acknowledging that a large list looks good to sponsors. He says the point of the newsletter is to share what he finds interesting and does for fun, to tinker and build random things, and to show readers they can do more with AI without going to the extreme of what he calls productivity-tokenmaxxers. He admits that making videos pushed him toward optimizing them and wondering whether he was trying to become a YouTuber, and says he will make them when it feels right rather than on a schedule.

He notes the irony of his position: he was previously the person telling people they did not need to learn to code, and now finds himself lost in multiple agents trying to optimize and automate everything in the belief it will make him more productive. His reframing is that an agent is a computer, not a person who needs to be kept busy and managed, and that it should work for him rather than the reverse.

For the next 30 days he plans to follow a short set of self-imposed rules: writing morning pages first thing to dump random thoughts and clear his head; using agents only to build things he finds interesting and fun; adopting no new tools unless they clearly solve a pain point; not summarizing long posts but either reading them fully or skipping them; and not doing things purely because they make money. He says he keeps thinking about packaging a "move fast and slow with AI" idea into a business but will spend much more time thinking it through before handing any of it to an agent, and will experiment across video, written and interactive formats depending on what gives him energy.

The post closes with a note thanking a reader named Ashe, who messaged him after a tweet and then spoke with him on a call, for reflections that helped him reconsider what he is doing. This is best read as a personal essay with a practitioner's observation attached rather than a product announcement; the main hard fact in it is the 60,000-plus subscriber removal, presented as a deliberate trade against sponsor-friendly numbers.

Why it matters: for developers and product teams, the piece is a signal that tool fatigue and agent sprawl are becoming a recognized problem even among people who cover the space professionally, which is inference on my part rather than a stated finding. Teams building agent products may face users who increasingly ask what pain a workflow actually removes, not just what it can automate. The 30-day rules are one practitioner's experiment, not a validated method, but they offer a concrete template for anyone weighing how many background agents and tools to keep running.

Read at Ben's Bites

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

Let’s think about what we’re doing here