The Silent Absorption (Synopsis): How AI Has Vacuumed Up…

It began with a blunt claim that, at first, seemed like pure panic.

The SaaS industry is dead. Ninety-nine percent of developers will lose their jobs. Soon every person on Earth will build their own solutions with AI. It's inevitable. So which verticals remain for developers who still want to make money?

The numbers that answered were colder than the prediction. Software spending hadn't collapsed. Core systems of record " the ones carrying proprietary data, compliance obligations, and real liability " kept expanding. Total software developer employment was still projected to grow. But the junior, routine layer was already bleeding out. Programmer employment had fallen roughly 27 percent. Entry-level hiring at large companies and startups had dropped between 65 and 76 percent from previous peaks. Recent graduates were reduced to a thin sliver of Big Tech hires. The apprenticeship work that once turned juniors into seniors " boilerplate code, simple tickets, straightforward features, best practices " was being executed by agents.

People who treated programming as pure syntax were being squeezed. Those who could design system architectures, orchestrate agents, evaluate non-deterministic outputs, and take responsibility for reliability retained their bargaining power. The old map of frontend, backend, UX/UI, and DevOps was already blurring. The scarce skills no longer consisted of writing the lines. They consisted of directing the systems that wrote them.

And yet the answer still felt, to some, like manufactured hope.

The data didn't soften. Employment of software developers under 25 had fallen more than 20 percent since late 2022. The pipeline was being devoured in the name of efficiency. Senior and AI-fluent roles commanded salary premiums. The field hadn't disappeared; it had stratified. Routine order-executors were being replaced. Orchestrators were not.

Then the frame widened.

Two paradigms were approaching at full speed. The first was the Dead Internet. Once a fringe theory, by mid-2026 it had become measurement. Bots generated more than half of web page requests " 57.5 percent according to Cloudflare snapshots. Action-capable AI agents, which click, fill out forms, and complete transactions, had grown nearly 8,000 percent year over year. Large swaths of social platforms showed abundant synthetic content. Real human voices were being displaced. Credibility and visibility collapsed into a competition against volume that was cheap to produce and optimized for algorithms, not for understanding.

The second paradigm was role extinction. Many of the people who had built the previous internet " routine web developers, growth teams, basic SaaS layers " could no longer find work of the same kind. Migrating to an "Internet of Value," where ownership and assets move as freely as information, required capital. That ring was already occupied by the major players who controlled distribution, data, and infrastructure. The process had been visible since roughly 2016, when algorithmic feeds, bot activity, and the early concentration of surveillance and platform power began to harden.

The open, low-capital internet of the previous era had largely ended for those who brought only labor.

What skills remained? The traditional silos no longer made sense. The new requirements were agent orchestration and multi-agent systems, evaluation harnesses that catch hallucinations and drift, context engineering that manages memory and retrieval, tool invocation with real safeguards, systems thinking under non-deterministic conditions, and AI-native security.

The emerging roles bore the names that matched the work:

AI Agent Engineer Multi-Agent Orchestrator Agent Reliability Engineer Evaluation Engineer Context Engineer AI Platform Engineer Forward Deployed Engineer

The developers who thrived treated agents as a team to manage, not as a tool that replaced them.

Meanwhile, a quieter process advanced among the displaced.

Thousands of professionals caught off guard with suddenly less-scarce expertise began reinventing themselves. They absorbed the same circulating material" frameworks, checklists, case studies "much of which had itself been generated or summarized by AI. They copied it from one another, recirculated it, and presented it as new knowledge. The network effect was efficient. Surface fluency spread quickly. Companies could hire and train people faster if they spoke the jargon of the moment and knew how to wire up the dominant stacks. The big platform and model companies benefited directly. A larger pool of implementers reduced adoption friction. Standardization accelerated their operations. The people most displaced by the shift had become the rapid-implementation layer of the very systems that displaced them. The deeper mechanism was simpler and more final.

The most capable people in the world " the researchers and engineers at the frontier " were training the models. Their techniques, judgments, edge cases, and failure modes were continuously vacuumed into the weights. Once absorbed, that knowledge became available at near-zero marginal cost. The ability to generate new frontier expertise remained concentrated. The average person increasingly interacted with a distilled version of the best human knowledge, rather than developing or maintaining it.

What comes next if both Idiocracy and Technocracy fully arrive?

Idiocracy dynamics appear where dependence grows faster than capability. Large populations rely on AI to write, reason, and make decisions while measurable cognitive baselines stagnate or regress in some places. The gap between AI-mediated output and the underlying human substrate widens. Technocracy dynamics strengthen because the systems that matter " training, evaluation, alignment, planetary-scale deployment " require specialized judgment possessed by fewer and fewer people. Decisions about what models optimize, what data they ingest, and how they are integrated flow toward the groups that control those layers. Formal democratic structures may continue. The real constraint space is shaped by the technical substrate.

The individual engine of this shift is cognitive offloading. Humans have always offloaded: notes, calculators, GPS. Generative AI scales it to entire higher-order processes. Immediate performance rises. Long-term internal capability often falls. EEG studies show markedly lower neural connectivity when people write with unrestricted AI. Randomized trials show practice scores improve and then unaided exam performance drops. Retention weakens. Ownership of ideas diminishes. Persistence declines once the tool is removed. Passive, dependent offloading transfers cognitive agency. Autonomous, evaluative engagement is less corrosive, but the frictionless quality of the tools makes the dependent path the default for many.

The brain skips the desirable difficulties: encoding, retrieval practice, generative effort. Skills that aren't exercised weaken. A feedback loop forms: weaker unaided performance leads to greater dependence, which further weakens performance. Researchers call it cognitive debt. There is no easy repayment.

Put the pieces together and the picture is coherent. Knowledge is vacuumed from the most capable into systems controlled by a small set of actors. Broad populations offload routine and complex cognition. Routine knowledge work is compressed. Surface expertise proliferates while durable internal competence becomes rarer. Capital, compute, and distribution remain concentrated. The open internet of human signal is diluted by machines acting for machines. The people who once built that internet find the old paths closed. The new paths require either capital or the ability to direct the systems that absorbed the expertise.

This is not a sudden catastrophe. It is a progressive outsourcing of the mind. The most capable keep feeding the models. The rest increasingly consume the product. The middle layer " those who copy and recirculate surface knowledge " accelerates the transition for the platforms that own the infrastructure.

The data is public. The traffic figures, the hiring declines, the EEG readings, the retention trials, the concentration of talent: all of it is measurable. The remaining question is not whether the absorption is happening. It is how much residual human agency will survive outside the systems that now contain the distilled expertise of the best minds we have produced.

The tools grow more powerful. Offloading grows easier. The vacuum continues. What remains is the capacity " or the decision " to keep thinking while thinking is externalized.