All the current LLM coding agents build on advanced anti-bot measure evasion features that were initially designed by humans to solve the insane "grab all the things" problem LLM models have had. Now, nothing is safe on the Web, and you don't know what half-assed vibe-coded application will pummel senseless the websites you host for little to no reason as long as tokens run.

A few years ago I used to find it cute in Cyberpunk 2077 that the Internet fell to rogue AIs and that the only way to be safe was local-only networks. Seeing the scale and the race to add more computing power towards an ill-defined AI infrastructure wrecking havoc on all public web spaces, I'm not so sure anymore.

What if Google/Meta/OpenAI simply don't want to pull the plug if one of their creation finally glasses the open Web because they are afraid of dropping their share price? Corporations' action are so little connected to the real world it sounds like a plausible scenario.

Now I can clearly envision a devastated public Web where uncountable AI agents create and consume each other's nonsensical websites faster than any human can react to, while parallel whitelisted meshes would develop out of necessity.

Making a Hidden Door Status Sensor

The door sensor in its new enclosures. (Credit: Dillan Stock)
A common sight in ‘smart homes’, door sensors allow you to detect whether a door is closed or open, enabling the triggering of specific events. Unfortunately, most solutions for these sensors are relatively bulky and hard to miss, making them a bit of a eyesore. This was the case for [Dillan Stock] as well, who decided that he could definitely have a smart home, yet not have warts sticking out on every single doorframe and door. There’s also a video version of the linked blog post.

These door sensors tend to be very simple devices, usually just a magnet and a reed relay, the latter signaling a status change to the wireless transmitter or transceiver. Although [Dillan] had come across recessed door sensors before, like a Z-wave-based unit from Aeotec, this was a very poorly designed product with serious reliability issues.

That’s when [Dillan] realized that he could simply take the PCB from one of the Aqara T1 door sensors that he already had and stuff them into a similar 20 mm diameter form factor as that dodgy sensor unit. Basically this just stuffs the magnet and PCB from an existing wart-style sensor into a recessed form factor, making it a very straightforward hack, that only requires printing the housings for the Aqara T1 sensor and some intimate time between the door and a drill.

youtube.com/embed/XVaGANL2T7o?…

hackaday.com/2026/02/10/making…

Building a Self-Playing Chess Board Robot

As popular as the game of chess is, it has one massive flaw. This being that it requires two participants, which can be a challenge. Although playing chess on a computer against an AI has been a thing for many decades, it’s hard to beat physical chess boards that give you all the tactile pleasure of handling and moving pieces, yet merging the two is tricky. You can either tell the player to also move the opponent’s pieces, or use a mechanism to do so yourself, which [Joshua Stanley] recently demonstrated in a video.

There are a few ways that you can go about having the computer move and detect the pieces. Here [Joshua] chose to use Hall magnetic sensors to detect the magnets that are embedded in the 3D printed chess pieces as well as their absence. These sensors are mounted to the back side of a PCB which is also the playing field, thus using the silkscreen for the board markings.

For the electromagnet that moves the chess pieces core x/y kinematics were used to move it underneath the PCB, engaging when moving pieces but otherwise deactivated. This is all controlled by an ESP32 MCU, while the computer runs the open-source Stockfish chess engine. As the human player changes piece positions this is detected by the magnet’s presence, with the change input into Stockfish.

As the demonstration at the end of the video shows, it definitely works, yet some issues remain. Ignoring the mistake with making the near-right corners black instead of white, the pieces are large enough that e.g. moving a knight piece between others pushes them to the side, requiring these to be put back in place.

There is also no way for the computer to detect which piece is placed where, which can be incredibly helpful on some commercial self-playing chess boards like this for new players, as well as to detect invalid moves, but this might be on the list for a potential V2 of this build.

Best part of this build is probably the use of a PCB for the playing field, which would allow you to go pretty crazy with custom designs and colors, especially now that some PCB places are offering multi-color silkscreens that allow for custom graphics.

youtube.com/embed/tLgXvUgsYmw?…

hackaday.com/2026/02/10/buildi…