Summarize Paper logo SummarizePaper.com

← Computing beyond AI this week · 9 November 2026 · 5 min read

Computing beyond AI this week: What noise reveals, and what guarantees conceal

New preprints explore the information hidden in aircraft noise, the limits of quantum and economic advantage, and why apparently good software and energy forecasts can disappoint in practice.

How to check the findings and keep reading? Illustrated guide

How these papers were chosen among the week's 1904 papers

All 1904 papers of the week whose main category is in this field were read, title and start of the abstract, and 76 were kept. They were then ordered by what can be counted: how many research profiles of our readers they come close to, whether a rising topic of the field includes them, and how many categories they were posted in. The abstracts of the first 40 were read in full to choose these, no more than two from one category and one per team.

1 of them belongs to a rising topic. The 14 numbers of the text were checked against the abstracts. 22 statements on what the papers found were looked for word for word in their abstracts.

Sources: arXiv.

An aircraft overhead can become a location clue #

An outdoor recording can contain more than its maker intended. AircraftWitness tests whether the changing pitch of passing aircraft, matched against public flight records, can reveal where a microphone was placed. In a preregistered test at 24 new places, the best-performing locator placed clear-pass recordings within 1 km at 4 places, compared with 0.57 expected from uniform guessing. Wrong timestamps and swapped recordings produced no successful locations.

The more revealing result may be how rarely the technique worked on ordinary material. Across 1,600 randomly selected public recordings, the locators returned answers at most 20 times and located at most 2 correctly. This preprint therefore identifies a real but conditional privacy exposure, rather than a general-purpose audio tracking system. The released benchmark and aircraft-pass labels should help researchers distinguish convincing demonstrations from practical prevalence.

Who works on it? Literature review Follow this subject

Security scanners can improve their scores without improving discrimination #

Does a security scanner configured today remain useful as software changes? A study of 20,000 vulnerable and fixed C/C++ cases compared releases of CodeQL and Infer alongside a Semgrep configuration. The researchers examined yearly populations, matched projects across years, and traced selected performance drops through repair edits and analysis logs. Newer CodeQL releases achieved higher F1 scores, but recall and false-positive rates rose together; a separate measure of discrimination remained near chance.

The investigation exposes a measurement problem as much as a software problem. Many examined alerts did not touch the lines changed by a repair, while Semgrep reported successful runs even when parser or timeout diagnostics were present. A tool can appear to improve by issuing more warnings, or appear to deteriorate because its analysis coverage changes. For anyone procuring or evaluating security software, the lesson is to ask what an alert actually demonstrates—not merely whether a headline score went up.

Who works on it? Literature review Follow this subject

A proposed boundary on unstructured quantum advantage #

A theoretical preprint reports a proof of the Aaronson–Ambainis conjecture, connecting the mathematics of simple polynomials to the power of quantum computation. In accessible terms, the conjecture says that a bounded, low-degree polynomial on binary inputs cannot spread all its meaningful variation so thinly that no individual input matters. The author gives an explicit lower bound on the influence of some input variable, using a proof described as elementary.

The computational consequence concerns query complexity: how many times an algorithm must inspect its input, rather than its full running time. The result implies that classical algorithms can approximate the acceptance probabilities of quantum algorithms on most inputs using a polynomial number of queries. If confirmed, it would sharpen the case that superpolynomial quantum query advantages require special structure, such as restrictions on which inputs are allowed—not settle the practical capabilities of quantum computers across the board.

Who works on it? Literature review Follow this subject

For the grid, uncertainty needs a timeline #

Power forecasts for computing facilities must describe not just how uncertain demand is, but how it changes from moment to moment. A benchmark built from measured power at a 448-GPU cluster follows forecasts through uncertainty calibration, generation of possible demand trajectories, and reserve procurement. After utilization changed, a nominal 90% prediction interval covered only 63.4% of observations; online calibration raised coverage to 87.3%.

Yet better coverage was not enough. Sampling each future demand independently generated unrealistic power ramps and encouraged excessive reserve purchases. Reconstructing temporal relationships from historical forecast errors lowered realized reserve cost by 7.4-10.6%. The broader lesson extends beyond data centers: a forecast can describe individual outcomes reasonably well while still misrepresenting the sequences that infrastructure operators must survive.

Who works on it? Literature review Follow this subject

Which bit best survives noise? #

Another theoretical preprint reports a resolution of the Courtade–Kumar conjecture. Begin with a uniformly random string of bits, then independently flip each bit with the same probability. If the noisy string must be reduced to a single bit, the claimed result says that no function of the whole string preserves more information about the original than simply retaining an individual noisy bit.

The abstract states the result for every allowed flip probability. Its importance lies in making a basic intuition precise: under this symmetric noise model, elaborate processing cannot outperform the simplest possible choice for this particular information measure. This is a mathematical statement about compression and information, not a recommendation that real communication systems should dispense with error correction.

Who works on it? Literature review Follow this subject

Randomness has a price ceiling in truthful scheduling #

Scheduling work across machines becomes harder when the machines' owners may misreport their costs. A mechanism must both distribute tasks efficiently and make honesty worthwhile. A new theoretical preprint establishes a tight lower bound for truthful fractional scheduling, where tasks can be divided, and transfers that bound to randomized mechanisms whose incentives reward truthfulness in expectation.

A complementary construction reaches the same leading constant for randomized scheduling. Together, the results say that randomness can improve on the optimal deterministic approximation by an asymptotic factor of two, but cannot remove the fundamental growth of the worst-case penalty as the number of machines increases. Beyond scheduling, this illustrates a recurring tension in market design: incentives constrain efficiency even when a planner can randomize.

Who works on it? Literature review Follow this subject

These preprints share a concern with what information and guarantees really buy us. A passing aircraft sometimes reveals a location; a security alert need not identify a repaired flaw; a calibrated forecast can still mislead an operator. The theoretical papers pursue the complementary question: where can rigorous limits replace intuition? Across both kinds of work, the useful advance is a clearer boundary between a promising signal and a justified conclusion.

Rising in this field

All the rising topics →

Ask all the papers of this letter Save the papers in a collection RSS

Was this letter useful to you?

Written by AI from the 2690 papers of the week in this field. Most are preprints, not yet peer reviewed: the text says what their authors report.

Get one letter by email every week with a free account, all of them with Premium. Create a free account