Benchmark in Milliseconds

How long should a micro benchmark run? My rule of thumb is to tweak the input size until the benchmark takes about 300ms, for the following reasons:

Subquadratic 3SUM and Subcubic APSP

Abstract page for arXiv paper 2610.06783: Truly Subquadratic 3SUM and Truly Subcubic APSP via Triangles in Sparse Lopsided Graphs

Show HN: I turned my iPhone and a $20 smart plug into an f-stop timer

A darkroom timer for iPhone and iPad.

Mathematics of Geothermal Energy

<p>The mathematics of geothermal energy involves the application of various mathematical methods to explore and optimize how geothermal heat from the Earth's interior can be harnessed for energy production and heating. Geothermal energy is derived from the heat generated by the Earth's formation and ongoing radioactive decay, with the temperature increasing significantly with depth, a concept known as the geothermal gradient. High-gradient areas, particularly along tectonic plate boundaries like the Ring of Fire, are prime locations for geothermal energy projects.</p> <p>Mathematicians and engineers utilize models such as Lagrangian–Eulerian flow models to understand fluid movement and its implications for geothermal reservoirs, while stoch

The Early History of Smalltalk (1993)

Tapo (Rust/Python library) now speaks TP-Link's TPAP protocol

tapo now logs in over the TPAP protocol, so the Third-Party Compatibility switch in the Tapo app can stay off. Also new: H200 and H500 camera hubs with recording downloads, and plug schedules and timers.

Gleam doesn't compile to Erlang source anymore

News post: Gleam v1.19.0 released!

Nature's capacity to 'bounce back' when species are lost is vastly overestimated

The idea that ecosystems carry plenty of backup, so if a small number of species become extinct there is little harm, has been greatly overestimated, according to the largest global study of its kind.

Find the flattest route between any two points in SF

flattensf: the flattest walking or cycling route between any two places in San Francisco, and every route between it and the shortest one.

Dust: Pretraining Transformers Without Backpropagation

Release of Polars 2.0

DataFrames for the new era

JetBrains reported a net financial loss first time in its tracked history

JetBrains (Media & Entertainment, Czechia): total revenue CZK 16,008 mil, net profit CZK -315 mil in 2025, ROE -11.1%. 721 indicators from filings, financials, KPIs and Excel download.

Meta's Muse Is an Adorable Privacy and Security Dumpster Fire

Meta's agentic AI product Muse has had a rocky few weeks since launch. The product, which features an animated avatar named Jolly (one presumes to make mass hyper surveillance seem adorable) is supposed to help you offload busywork like making restaurant reservations, paying bills, or ordering groceries. Despite Meta having claimed repeatedly that Muse was…

Utah to let AI examine patients and prescribe medication without human oversight

The good news is that the program is currently very limited in its scope. It's just one year long and only for patients with mild-to-moderate acne.

World's First enhanced geothermal power plant completed in just 23 months

Fervo Energy completed its first power plant in less than two years. The next phases promise to connect to the grid even quicker.

Nobel Prize in Physics goes to Francis Halzen

The Nobel Prize in Physics 2026 was awarded to Francis Halzen “for decisive contributions to the IceCube Neutrino Observatory and the discovery of high-energy neutrinos of astrophysical origin.”

Mistral Large 4

The most powerful AI platform for enterprises. Customize, fine-tune, and deploy AI assistants, autonomous agents, and multimodal AI with open models.

Beam: Reflection's 501B open-weight model

Make intelligence open and accessible to all

AI is now capable of developing its own inference hardware

An open-source AI accelerator, developed by AI: RTL, ISA, simulator, compiler and profiler in one repo. Runs Qwen3, LFM2.5 and Qwen3.5 on a Kintex-7 PCIe card. - FeSens/openTPU

Show HN: Parseable, an open observability datalake, handles 100M time-series/min

Unified observability platform for logs, metrics, and traces. AI-native insights, S3-native storage, SQL queries, OpenTelemetry-native, and 80x lower cost than Datadog and Splunk.