David Brin's Hypersecretary and the AI We Never Built

Earth, by David Brin

In 1990, in his novel Earth, David Brin imagined a personal AI called the Hypersecretary. The Hypersecretary would manage your interaction with an overwhelming global network, spawning task-specific sub-agents to fetch information, verify facts and negotiate access. It could ingest and comprehend meaning rather than simple search words. It filtered ruthlessly but injected deliberate randomness to prevent you becoming trapped in a bubble of your own preferences. It drafted your responses but never sent them without your review. For context, and to save you from looking it up: in real life, the first search engine (at the time, pre-web), Archie, was also released in 1990. The better-known early web search sites - Yahoo, AltaVista - emerged in the mid-90s. Google.com was registered as a site in 1997.

Over the last couple of years, parts of Brin’s vision have finally become reality, through technologies built on top of large language models. Agentic AI and agentic personal assistant systems, such as OpenClaw and Hermes, can perform many of the functions Brin envisaged. But Brin’s version had a feature ours didn’t inherit: it was obsessive about the truth.

I read Earth when it came out. I didn’t exactly love it. To be honest, I thought it an overstuffed novel that never really achieves take-off velocity, and contemporary reviewers fairly called it “a thesis in literary form.” If you want to enjoy Brin, start with the first Uplift trilogy (Sundiver, Startide Rising, The Uplift War). He has other good novels as well, for example the underrated Kiln People. But I still keep thinking about Earth, because what Brin got right, and wrong, says something about our own blind spots on AI.

David Brin is a professional physicist, with a PhD in Space Science and a background in optics and astrophysics. His extrapolations were and are informed by genuine expertise. He borrowed his fragmented, information-saturated narrative style from John Brunner’s Stand on Zanzibar (1968), but where Brunner immersed readers in suffocating overload, Brin proposed a solution.

My career has been that of a software architect. Reading Earth from a professional perspective, Brin’s Hypersecretary reads convincingly, even now. It’s not a single monolithic program but an orchestrator that spawns and manages lightweight agents (called “zesties”). Each agent has its own compute budget and a predetermined lifespan; when credits run out, it dies. The agents leave data trails that other agents can follow. The system’s swarm intelligence thus emerges from its simple components: if a thousand agents from different users are hunting the same information, their trails converge on reliable sources. So an individual agent just has to follow the network’s herd intelligence.

The human user stays in the loop to provide governance. One of the novel’s characters dictates her responses and lets the system clean up her syntax, but never sends anything until her designated review days.

Humans can also configure their systems with filters, which determine their behaviours. That same character sets her system to surface 20% of items at random, thereby deliberately defeating her own parameters. An AI that only tells you what you want to hear is, in Brin’s recurring critique of modern media, lobotomising its user. He anticipated denial-of-service attacks, which he called “chaffing” - whereby adversaries would try to flood specific parameters with truth-like noise until the system’s certainty drops so low it stops reporting.

Brin even posited that these filters could be inherited as financial assets. In Earth, you can pass down your “Filter Libraries” - refined sets of trusted sources and exclusion rules accumulated over decades. A “Century-Old Filter” is described as more valuable than land; this, in turn, creates a digital aristocracy that develops across generations, much like Western plutocracies clustered around land wealth. Well, we don’t have inheritable attention infrastructure yet. But you can see where he was going with this, one grand idea among many.

His agents run on a “vouching” system - a chain of who had staked their reputation on the data’s accuracy - this assigned value both to the data and to its underwriter. If a high-reputation source vouched for a lie, their credibility took a measurable hit across the entire network - a chain of truth, where signatory authority is a liquid asset that depletes when you’re wrong.

This verification mechanism goes even deeper. Agents run a “Liar’s Paradox” test: any data deemed to be “too perfect” generates a negative flag. The agents actively search for dissent. Absence of dissent is suspicious; in a healthy information ecology, there should always be some noise and disagreement. Absolute consensus is a sign of a hacked network.

Here in the real world, we’ve built some of this architecture, but I wonder whether we’ve built the right bits. A technical reader skimming this can see echoes of blockchain architecture, for example. And Google Maps infers traffic congestion from phones reporting their location; that’s an example of sensor-grounding. But we didn’t extend these principles to claims that actually need verification. Nobody’s cross-referencing news reports against physical sensors. Nobody’s running Liar’s Paradox tests on search results, except maybe a large number of extremely frustrated end users.

This absence arises from the fact that modern technology companies took one particular set of programmes, designed for search interactions and to convincingly masquerade as human sentence-creators, and then applied them to just about everything else. We built fluent bullshitters with no native concept of truth. When an LLM confidently cites a paper that doesn’t exist, or comes up with some other falsehood that it insists is true, it’s referred to as a “hallucination”. But these behaviours arise because the system’s underlying capabilities were primarily designed to convince the user, not to represent underlying and fundamental notions of what is true.

Brin’s Hypersecretary is specifically designed to serve its user, not to enable a corporation to monetise the interaction. One character hires a “rogue hacker” to build hers because she suspects the off-the-shelf versions have corporate or government backdoors. In Brin’s imagined world, the user always has an alternative: you could choose to ignore the standard version and pick a more trustworthy builder instead. Ultimately, the Hypersecretary works for you; in the classic words of Tron, it “fights for the users”.

Brin’s vouching architecture isn’t implementable without additional protection layers - bad actors can potentially game the reputation systems. His counter is to open-source the algorithms. If you can see how reputation is calculated, you can spot whether or not the gaming is happening. It’s an answer, though not a complete one (you might be able to spot it, but such things always have edge cases, and it’s at the edge cases that exploits tend to happen). His instinct, however, is correct: provenance matters more than fluency. Compare with our current situation, where LLMs are bolting on citation and retrieval after the fact, retrofitting truth onto systems primarily designed for plausibility.

Brin compares our current moment to the 1930s, when radio and loudspeakers amplified demagogues and nearly wrecked civilisation. Every information transition produces this crisis; the pessimists are “currently right,” he argues, because we haven’t yet built the transition tools. The Hypersecretary was supposed to be one of them.

Brin’s 1990 vision assumed AI would be trustworthy but overwhelming. The user would be flooded with “good” data. The user’s problem was finding the signal in the noise, managing attention, and not disappearing into their own beliefs. What we got was slightly different: fluent and highly believable but untrustworthy. The user’s problem, therefore, is knowing whether to believe what the agent tells them.

Brin views science fiction as “self-preventing prophecy” - stories designed to frighten people into action. He places Earth alongside 1984, Silent Spring, and Soylent Green. He’s “often accused of being an optimist,” he says, but sees only a “60% chance we’ll eke through” - barely good enough odds to justify having children. Of course, he also had and has something today’s AI companies lack: freedom to roam. No capital bets. No product to ship. No investors to satisfy. Brin can more easily afford to be wrong; current tech executives can’t, and that directs what they’re willing to say. When Sam Altman or Dario Amodei or Jensen Huang tells you where AI is going, they’re not disinterested observers. Their claims serve their interests, framing the future to make their current bets look prescient. They’re predicting and building at the same time, so that their predictions are entangled with their economic incentives in a way an SF author’s claims typically aren’t.

Brin’s example suggests a pattern: we’re good at extrapolating capabilities - the shape of the technology, the problems it will address - but bad at predicting political economy - who controls the agents, whose interests they serve and what happens when the tool works against you rather than for you. Experts are usually right when they can apply their expertise in a debate. The things they don’t talk about, because they never considered them, are where the surprises emerge.

That blind spot - political economy - brings us to the prediction that troubled me most in 1990. It’s about what happens when an information ideology acquires geopolitical weight. In Brin’s 2038, there was a “Helvetian War” - a global coalition against Switzerland. The justification: Swiss banks hoarding wealth while the rest of the world suffered ecological collapse. The global network had made it clear that kleptocrats were hiding stolen assets behind secrecy laws. The war was framed - not as a conquest - but as forcing open the vaults to enforce transparency.

Brin’s narrative is sympathetic - secrecy is the ultimate sin in his moral framework, so transparency, it seems, is worth a war. In 1990, this seemed to me to be thoroughly absurd - a plot contrivance revealing the author’s ideological hobby-horses more than any plausible future. Surely the international order wouldn’t sanction a war for data transparency? Surely “might makes right” had guardrails? And it wasn’t even clear this was right in the first place.

It no longer feels absurd. We’re watching an American executive increasingly unconstrained by norms - extraterritorial application of law; weaponised financial infrastructure; forced divestiture of apps under national security framing; semiconductor export controls as capability containment. At this point, it’s coercion, not (yet) invasion, but it feels like the start of a path, not the final destination.

Science fiction has a way of normalising futures before they arrive - not prediction so much as rehearsal, where the story itself shapes what becomes thinkable. The tech brotherhood established a truth: data should be free, information wants to flow, secrecy is illegitimate friction. Brin’s novel established that opposition to such principles could form a valid casus belli. How soon before some weak country is pressured, overthrown, or occupied because it sits on vital rare earths, the best energy grid, the right latitude for a hyperscale data centre? On our current track, it doesn’t feel so far away.