Ollama on Apple silicon is now built on top of Apple’s machine learning framework, MLX, to take advantage of its unified memory architecture.

This results in a large speedup of Ollama on all Apple Silicon devices. On Apple’s M5, M5 Pro and M5 Max chips, Ollama leverages the new GPU Neural Accelerators to accelerate both time to first token (TTFT) and generation speed (tokens per second).

We’ve all been there. We learn ls, cd, and grep, and then we sort of… stop. The terminal becomes a place we live in-but we rarely bother to arrange the furniture. We accept that certain tasks take forty keystrokes, completely unaware that the shell authors solved our exact frustration sometime in 1989.

Here are some tricks that aren’t exactly secret, but aren’t always taught either. To keep the peace in our extended Unix family, I’ve split these into two camps: the universal tricks that work on almost any POSIX-ish shell (like sh on FreeBSD or ksh on OpenBSD), and the quality-of-life additions specific to interactive shells like Bash or Zsh.

GitHub stopped updating its status page with aggregate uptime numbers some time ago — if you use it regularly, you might have a feeling why. This is the missing version.

We rebuild platform‑wide and per‑service uptime from archived status updates, derive minute‑level downtime windows, and map incidents to services whenever the source data allows. Everything is open source, and PRs are very welcome!

FFprobe is a command-line utility that is used to gather information from multimedia streams. It extracts and prints the in-depth information about the stream’s format and individual coded frames. It can handle a variety of formats, including but not limited to MP4, MOV, AAC, MP3, and more.

Example Commands

Here are some example commands that demonstrate the power of FFprobe:

1. Basic Stream Information: The following command prints basic information about a multimedia stream:

   ffprobe -i input.mp4

   In this command, -i input.mp4 specifies the input file.

2. JSON Output: The following command prints the stream information in JSON format:

   ffprobe -v quiet -print_format json -show_format -show_streams input.mp4

   In this command, -v quiet suppresses the logging output, -print_format json specifies the output format, and -show_format -show_streams tells FFprobe to print information about the file format and individual streams.

In Linux, FreeBSD, and NetBSD, /dev/full, or the always-full device,^[1][2] is a special file that always returns the error code ENOSPC (meaning "No space left on device") on writing, and provides any number of zero bytes to a process that reads from it (similar to /dev/zero).^[3] This device is usually used when testing the behavior of a program when it encounters a "disk full" error.

$ echo "Hello, World" > /dev/full
bash: echo: write error: No space left on device

Despite widespread industry recommendations, a new ETH Zurich paper concludes that AGENTS.md files may often hinder AI coding agents. The researchers recommend omitting LLM-generated context files entirely and limiting human-written instructions to non-inferable details, such as highly specific tooling or custom build commands.

The team (Thibaud Gloaguen, Niels Mündler, Mark Müller, Veselin Raychev, Martin Vechev) justified the research by noting that while 60,000 open-source repositories currently contain context files such as AGENTS.md, and many agent frameworks feature built-in commands to auto-generate them, there has been no rigorous empirical investigation into whether these files actually improve an AI agent’s ability to resolve real-world coding tasks.

Everyone knew that an attack might be in the works—some American aircraft carriers had already been deployed to the Middle East weeks ago—but the Iranian government was caught off guard by the timing. Although the ayatollah surely was aware of the risks to his life, he presumably did not know that he would be targeted on this particular Saturday morning. Yet on Polymarket, plenty of warning signs pointed to an impending attack. The day before, 150 users bet at least $1,000 that the United States would strike Iran within the next 24 hours, according to a New York Times analysis. Until then, few people on the platform were betting that kind of money on an immediate attack.

It’s possible that artificial intelligence is something unique in human history, but the mass automation it seems bound to produce definitely isn’t.

Despite this, we rarely hear in any detail about previous waves of automation. There’s discussion of the Industrial Revolution, but that’s about it. We hear more about Engels’ Pause than we do about flagmen or telephone operators or motion picture projectionists.

This seems strange, because there has been a huge wave of automation within living memory. In fact, we are still living through it.

In the early 1970s disaster struck the Swiss watch industry. Now people call it the quartz crisis, but in fact it was a compound of three separate disasters that all happened at about the same time.