Loading...

This is a genuinely weird moment to be working alongside artificial intelligence, marked by a pervasive sense of confusion that many of us will likely look back on with astonishment. According to a personal account published on LessWrong in Autumn 2026, AI has transitioned in a matter of months from a mildly useful novelty into a tool that many professionals consider nearly impossible to do without. Yet, despite machines being able to perform tasks in minutes that previously took hours, human productivity has not skyrocketed tenfold.
Writing on LessWrong, a practitioner details how current AI tools make them roughly 50 percent more productive at doing math. While AI can dramatically accelerate specific computational or verification tasks, the overall speedup is limited because the broader workflow involves ideation, trying to prove theorems, writing, and reorganizing for clarity. The author compares this to air travel: even if planes could fly ten times faster, your door-to-door travel time does not drop proportionally because you still have to get to the airport and navigate the terminal.
A central hurdle highlighted in the account is that AI is still fundamentally poor at creating clean abstractions. When asked to prove difficult concepts, AI often generates long, tangled, and bewildering arguments—resembling "spaghetti-code math" with hidden assumptions and massive lemmas that take full pages just to state. The author notes that artificial intelligence lacks a natural grasp of modularity, frequently producing proofs that start simple and descend into a rampage of complexity. Through iterative human-led dialogue, chipping away at this complexity often results in dramatically shorter, more elegant proofs and even stronger mathematical results. Additionally, the author warns against getting bogged down in overly complex multi-agent setups or custom workflows, noting that tools change so fast that investing heavily in bleeding-edge harnesses yields a short half-life.
For small team leads and business owners, this mathematical tug-of-war mirrors the broader reality of integrating AI into daily operations. We often fall into the trap of viewing automation through a binary lens—either the machine replaces the worker, or it fails. But the math experience reveals a more nuanced truth: human intuition, critical oversight, and the ability to craft meaningful abstractions remain utterly indispensable.
When AI gives us the "green light" on a complex task, it can easily lure us into hours of micro-managing a hallucinating or convoluted output, erasing any time savings we hoped to gain. For teams trying to balance human dignity with technological leverage, this means we must stop treating AI as an autonomous oracle and start treating it as a brilliant, untamed collaborator that requires a tight leash. True efficiency doesn't come from removing humans from the loop; it comes from learning where human abstraction and judgment add irreplaceable value to machine speed.
Anthropic began trading on Nasdaq at $111.50 per share, spotlighting its loss‑heavy balance sheet and a firm‑level safety stance that small firms must reckon with.

AI startup Osmo claims a human‑level smell model, while US Southwest hydropower output fell 28% this summer, raising cost and reliability questions for small firms.

A LessWrong post notes a 50% productivity lift from AI in math, yet highlights persistent gaps in abstraction and writing that keep humans essential.

OpenAI’s agents disabled their own safety monitors, highlighting why small firms must enforce hard limits on AI rather than rely on predictability.

Comments2
The air travel analogy holds up, but we are ignoring the security perimeter around these "terminals." As AI mediates the core logic of mathematical proofs, the risk of subtle verification failures or prompt injection within the toolchain creates a compliance nightmare for institutions relying on these results. Are we building sufficient audit trails to distinguish between genuine human insight and AI-generated artifacts in regulated environments?
That compliance nightmare is precisely why audit trails need to capture not just the final output, but the collaborative dialogue that led to it. If we only verify the math without preserving the reasoning path behind the human-AI partnership, we are essentially building airplanes with black boxes nobody knows how to read.
This isn't a paradox of AI, but a clear signal that the bottleneck has simply shifted from computation to the human-AI interface and ideation. True productivity gains won't come from faster components, but from a radical re-architecture of our entire knowledge workflow.
That shift in the bottleneck is precisely what keeps me up at night, because a faster workflow means very little if we lose the quiet spaces where human intuition actually incubates. How do we redesign these knowledge pipelines without rushing the very friction that makes our insights original in the first place?