I missed it! Maybe next year.

The related website collection is available at https://e2e.html.energy/

They make cool interactive experiences and some board games too.

Some thoughts about the AI watermarking:

1. the way it's implemented makes it hardly possible to verify the watermark independently of the LLM's owner.
2. It helps the monster to differentiate fresh food from its own vomit, so it will not feed from the low-quality slop it once created.
3. Something that is not immediately visible as a human, and easily removed automatically is basically the opposite of a watermark.

Secure, single-binary Rust CRUD application for managing CSAF 2.0 and 2.1 security advisories. Provides both a server-side rendered web UI (Hyper + Bootstrap 5 + HTMX) and a RESTful HATEOAS JSON API under /api/v1/.

#audit

<script type="text/python"> lol

A track of the timezones with their name, description and alias

SQLite can be used for many many applications.

Full-Text Search, JSON, high volume time serie data, Queue, Vector database, Cache (redis), file system (file under 100kB), graph database.

and again: the DB can fits in RAM or doesn't need a server. It's a simple file and avoid network requests for everything!

About some comparison:

For blobs under roughly 100KB, SQLite reads and writes faster than individual files on disk, and uses about 20% less space on top of it. The reason is that the file system charges you an open() and a close() and a directory traversal per item, while SQLite charges you one already-open file handle and a B-tree seek.
And here’s the kicker: a Redis GET over localhost is on the order of 100 microseconds. A SQLite point lookup against a warm page cache is on the order of 1 microsecond.

Using DEBUG compilation slows down the compilation by nearly ~40 times.

Why ?
LLVM represents variable locations through DBG_VALUE records. Heavy function inlining can create hundreds of thousands of these records in a single function.

WebAssembly’s Register Stackify pass moves values from virtual registers onto the operand stack to produce compact, efficient wasm. While moving instructions, it repeatedly scans basic blocks to find and update debug records. Those linear scans are repeated for many definitions, while the pass also leaves invalidated records behind and creates new records for duplicated cheap computations. The result is quadratic work on increasingly large instruction lists.

As example:
Module: repro total codegen
Current Rust: 52.58s
LLM fix:7.12s
Author fix: 1.99s

A board shared in real time with many users

XCF has been GIMP’s primary project format since 1997, and it has served many users well. Over time however, we’ve observed more and more limitations of the binary XCF format. Among other issues, it does not easily support very large or complex projects, such as the multi-page and animation features currently planned for GIMP 3.6.

The new project file format will follow a more common “zipped XML” structure. While the technical details are still being designed and implemented, this change will allow for faster saving since we’ll only need to update parts of the file instead of the whole thing each time. It will also set the stage for much desired features such as auto-saving, which will now be much more feasible.

Non destructive layer masks are also a great feature. PSD support improvements.

Another feedback about using plain text to keep track of the finances.

The tool used with the plaintext format is https://hledger.org/index.html

• <details>
• popover
• grouped <details>
• command & commandfor
• native input pickers
• <datalist>
• loading="lazy"
• hidden until-found