L2

You can project n points into O(log n / eps^2) dimensions and keep every distance. The theorem underneath every vector database.

Erasing one bit costs kT ln 2 joules, unavoidably. The point where information theory becomes physics.

PAC learning and VC dimension's bounds predict that an overparameterized network cannot generalize, and it generalizes anyway.

Averaged over all problems under a uniform measure — a distribution no real task is drawn from — every learner ties.

Error splits into three terms, one of which is irreducible. Still the right first diagnosis of a bad model.

No ranked voting system satisfies a short list of obviously reasonable conditions. Applies directly to prioritization, ranking, and consensus UX.

Every optimization problem has a shadow problem whose optimum is the same number, and it certifies your answer: the only post in the series about proving your...

The bottleneck equals the capacity, exactly. A duality result you can see, and the workhorse behind matching and segmentation.

The constant factors that dominate asymptotics in practice, and cache-oblivious algorithms as the theoretical response.

The unimplementable best cache policy, and the proof that LRU is k-competitive with it. The application of Competitive analysis and ski rental, so it follows it.

A compiler can hide a backdoor in itself with no trace in any source. The limit of source-code review, and what reproducible builds do about it.

Rate limiting has a theory: bound the arrival curve and you get a provable bound on delay and queue depth.

The M/M/1 queue's 1/(1-rho) blowup. Why a service at 80% load is fine and at 95% is on fire.

All of modern crypto rests on an unproven assumption. Why P != NP would still not be enough to secure anything.

Security lives in the key; and Shannon's proof that the one-time pad is unbreakable and impractical for the same reason.

Your hardware and compiler reorder your program. What, fences, and acquire/release actually guarantee.

2PL produces exactly the serializable schedules; anomalies are precisely what you buy by weakening it.

The correctness condition for concurrent objects, and how it differs from serializability (Conflict serializability and two-phase locking) and sequential...

Peterson's algorithm and Lamport's bakery. Correct mutual exclusion from plain reads and writes, and why modern hardware still breaks it (see Memory models and...

A randomized scheduler with a provable bound on both time and space, and why every modern runtime converged on it.