In Brief
Illustration Beyond Proof
Why this matters now: Terry Tao’s essay on exposition in mathematics warns that AI‑generated and machine‑formalized proofs risk producing technically correct but unreadable mathematics, with implications for researchers, students, and tool-builders today.
"Reject irresponsible and unsustainable usages of AI technology that only serve to advance nominal goals rather than the true underlying goals of the field," writes Terry Tao, who also finds it "disturbing to see proofs ‘abandoned at an intermediate stage’."
The original post argues that mathematical value lives as much in intuition, examples and exposition as in a line‑by‑line formal verification. With companies and research groups publishing mass-produced proofs and formalizations in systems like Lean, Tao urges the community to preserve the practices that make results usable: talks, worked examples, motivation, and human-friendly narratives that let others build on an idea.
Whatever became of holography?
Why this matters now: A retrospective on holography traces how a once‑flashy consumer dream settled into steady, valuable niche uses — a useful case study for tempering hype around emerging optics and display tech.
"Although it came to represent high technology, holography was born from unlikely sources," notes the Americanscientist feature.
The piece shows holography did not disappear; it quietly became a dependable tool in security, metrology, microscopy and specialized imaging. The public spectacle receded, but underlying physics, new computational tricks, and materials work keep it relevant to AR optics and advanced imaging — a reminder that practical maturation often looks less like a boom and more like steady engineering.
Deep Dive
Illustration Beyond Proof
Why this matters now: Terry Tao’s concerns about AI‑assisted and machine-verified proofs are immediate for any mathematician relying on or producing formalizations, because sloppy exposition can turn correct results into dead ends for human readers.
Tao’s essay sits at the crossroads of two fast-moving trends: powerful generative models that can draft proofs, and growing efforts to formalize mathematics in proof assistants. Both reduce friction in producing syntactically correct arguments. The danger Tao highlights is cultural: if the community prizes formal correctness alone, we risk accumulating a library of proofs that nobody understands intuitively, that provide no route map for generalization, and that fail to teach students how to think.
A few concrete frictions make the worry real. Formal systems force a level of detail that obscures the high‑level ideas; AI can generate plausible but brittle intermediate steps; and publication incentives sometimes reward novelty or formalization counts over careful exposition. Tao calls for a norm shift: not just releasing formal proofs or machine output, but pairing them with human‑oriented explanations, examples, and follow‑up work that turns a proof into usable knowledge.
Practically, the community can respond on three fronts. First, reviewers and journals should evaluate exposition as a first‑class part of correctness — ask not only "Is this true?" but "Can a motivated reader reuse this idea?" Second, tooling should integrate human narratives with formal artifacts: literate programming for proofs, interactive visualizations, and annotated Lean scripts that embed the "why" alongside the formal lines. Third, training matters: graduate programs should treat exposition and teaching as core duties, not extras, and hiring committees should reward communicators as much as formalizers.
These fixes are feasible but nontrivial. They require modest changes in incentives (conference requirements, reviewer guidelines), investment in tooling that lowers the friction of writing readable formalizations, and cultural leadership from senior mathematicians and institutions. The payoff is high: formal verification becomes not just a trophy but a foundation that accelerates further discovery because humans can read, teach and extend the work.
Closing Thought
If verification scales faster than explanation, we risk filling the shelves with technical trophies rather than tools. Terry Tao’s warning is simple and practical: preserve the human‑facing craft of explanation even as we adopt machine assistance. That doesn’t mean rejecting formal tools — it means insisting that every formal artifact carries its story, examples, and road map for the next person who wants to use it.