AI FOMO, Part 2: Attention, Wealth, and Professional Relevance

All English wording below is an editorial translation/paraphrase of Italian auto-generated captions; nothing is a verbatim quotation.

Salvatore Sanfilippo follows his first FOMO video by separating three anxieties: not running enough agent sessions, missing an opportunity to become wealthy, and becoming professionally irrelevant. His strongest practical claim is that visible agent activity is not the same as useful output. While people still supply product direction and integrate results, attention remains the bottleneck.

The workflow recommendations, project reports, business advice, and employment forecasts are Sanfilippo's judgments rather than measured comparisons. The recording provides no tasks, outputs, labor data, or economic analysis with which to generalize them.

Attention, Not Session Count

• Many terminals can reduce the quality of the human contribution: Sanfilippo says constant context switching makes it harder to remember each project's state, notice details, and steer work with care. He recommends two or three simultaneous projects, preferably in the same area, rather than dozens of interactive sessions (00:05:58–00:07:01(opens in a new tab)).
• Short-horizon work needs focused interaction: When a task requires rapid back-and-forth, he recommends making it the active task instead of repeatedly rotating through other sessions. This is a personal operating rule, not a tested concurrency limit (00:12:12–00:13:11(opens in a new tab)).
• Long-horizon investigations are the exception: He describes assigning an open-ended investigation for a day or two and asking the agent to notify him only when it finds a meaningful result. The example shows asynchronous delegation, but the video does not evaluate result quality or address the permission and secret-management risks of giving an agent messaging credentials (00:07:01–00:09:07(opens in a new tab)).
• Judge throughput by outcomes: As a counter to social-media demonstrations with many open terminals, he asks what projects, companies, or other results the workflow produced. The question is a useful evaluation prompt; the recording does not supply a comparative sample (00:13:11–00:14:14(opens in a new tab)).
• More autonomous systems may change the boundary later: Sanfilippo imagines research, feature ideation, implementation, and user-oriented QA operating as a continuous agent system. He believes part of that is already possible, but his present recommendation assumes that human product judgment still matters and compute is limited (00:10:02–00:12:12(opens in a new tab)).

Wealth and Professional Relevance

• Agent intensity and startup economics are different problems: If wealth is the objective, Sanfilippo recommends deliberately building a company around a problem customers, investors, and buyers care about. He warns against defining the company by a transient implementation technique such as RAG when the durable problem is access to information. This is strategic advice, not evidence that the proposed approach produces successful companies (00:14:14–00:17:04(opens in a new tab)).
• Professional displacement is a credible but uncertain concern: He expects routine implementation roles to face pressure while specialist technical knowledge and builder-like product ownership become more useful. He also expects job creation and displacement to coexist across sectors and periods. The video offers no employment data, causal model, or timeframe for those forecasts (00:18:02–00:21:57(opens in a new tab)).
• Individual overwork cannot solve a broad labor transition: Sanfilippo ultimately treats widespread displacement as a political and socioeconomic problem rather than something one developer can outrun by opening more sessions. His broader claims about future abundance, distribution, and AI energy use are opinions in this recording and are not supported with sources or measurements (00:21:57–00:25:03(opens in a new tab)).

Watch the first FOMO video(opens in a new tab), then watch this follow-up on YouTube(opens in a new tab).