Comments URL: https://news.ycombinator.com/item?id=50321048

Points: 136

# Comments: 7

Comments URL: https://news.ycombinator.com/item?id=50319876

Points: 8

# Comments: 2

Comments URL: https://news.ycombinator.com/item?id=50319662

Points: 6

# Comments: 1

Comments URL: https://news.ycombinator.com/item?id=50319563

Points: 143

# Comments: 43

As it turns out, that’s not so easy! I had a few ideas, but beyond the core mechanics, a game also needs to be interesting and easy to explain — after all, nobody (including me) wants to spend a long time reading through rules.

Let’s start with the concept, which came to me quite spontaneously: what if we represented letters like the digits on a digital clock — using lines, with dots at the points where those lines intersect? For instance, the letter "E" would consist of four lines and six dots — three at the intersections and three at the ends of the lines.

It turns out that every letter of the Latin alphabet can be represented this way (with a minor caveat) using three rows of dots connected by some lines. This immediately presented a challenge: how do you represent O-shaped parts, like those found in "B", "O", or "R"? I decided not to use O-shapes, an use onl rectangles/squares. This "alphabet" is always displayed below the game board, so the player can see the letters.

The goal of the game is simple: you are presented with three rows of dots and a set of lines, and you have to figure out the word. A great feature of the game is that it’s possible to deduce the word in a single turn — all the necessary information is right there on the board, allowing you to arrive at the answer with certainty. And what if you can’t figure it out on the first try? You can enter a guess, much like in Wordle: if the lines from the entered word are part of the target word, they turn green; if not, they turn red. The same applies to the dots. As you play, more and more lines gradually appear...

Why provide some lines at the start? To uniquely identify a specific word — or a small set of words — from the dictionary. For example, you could select a configuration of lines that corresponds to only one word, like "RADIO", or perhaps to just two possible words, such as "WEIGH" and "WRONG".

The rules seem clear, but the mechanics — not so much. There are dots on the board; when I enter a word, I realize there aren't enough dots. What should I do? Initially, if the target word had 5 letters, I allowed users to enter 4- or 5-letter guesses, but that just confused players. So, I came up with a different solution: adding red dots for dots that aren't part of the target word.

Another drawback is that the game is somewhat skewed. Since letters are processed from left to right, the first letters are revealed more frequently than the last ones.

Then there’s the game design itself: should the target words be 5 letters long, or 6? Should a single board allow for multiple possible answers (like "WEIGH" and "WRONG"), or should it be limited to one — or perhaps more than two? Should the difficulty increase from Monday to Sunday? Currently, it’s set up this way: Monday is the easiest (a few 3-letter words), while Sunday is the hardest (more than two 7-letter words).

The tech stack is simple: I use Flask for the backend and Python for the algorithms; I wrote the algorithms myself. While the JS, HTML, and CSS were created with the help of AI.

I’d love for you to test the game and share what you like, what you don’t, and what could be improved.

Specifically, I’d like to know:

1. Is the game and its goal clear?

2. Is it fun to play, and what could be improved?

3. Would you share it with your friends?

I welcome any constructive criticism.

You can find the game at https://www.jointter.com/

It’s free and requires no registration.

Thank you very much.

Comments URL: https://news.ycombinator.com/item?id=50317548

Points: 14

# Comments: 23

Comments URL: https://news.ycombinator.com/item?id=50316117

Points: 34

# Comments: 26

Comments URL: https://news.ycombinator.com/item?id=50316100

Points: 298

# Comments: 167

Entire inference engine is 510 bytes of MBR, weights, vocab, and strings reside in the sectors following MBR. Small char-level neural network was trained on dmesg generated by machine, so machine produces Linux boot messages infinitely.

train.py performs precisely the same integer arithmetic operations as the asm code does and thus produces an output identical character-by-character to the one produced by QEMU. The difference analysis revealed my favorite bug. The characters at the beginning matched those generated by the simulator, but then the output diverged. I retrieved the hidden layer via QEMU’s debug port and compared the neurons' values with those of the simulator, one neuron at a time. Only the value of the first neuron was correct. The activation involves idiv, and idiv stores its remainder in edx. The pointer to the biases table was stored in edx, which resulted in the first neuron being biased correctly and the following ones using garbage. Now the bias' address is determined from the loop counter, and no pointers have to survive the division. Another bug I found simply by looking at the code before I did my first run-through: the BIOS print routine zeros out bx, but bx contains the predicted character number. This explains the push and pop of bx surrounding the print.

sectorllm is the most similar to my project, as it runs an actual 5-layer llama2 with attention mechanism in just 1.2K of assembly code. It is the trade-off in the opposite direction, as my engine needs to fit into one sector and therefore my network has just one hidden layer of int8 operations for dreaming the dmesg.

Comments URL: https://news.ycombinator.com/item?id=50315782

Points: 88

# Comments: 5

Comments URL: https://news.ycombinator.com/item?id=50315217

Points: 9

# Comments: 3

Comments URL: https://news.ycombinator.com/item?id=50314617

Points: 28

# Comments: 7

Comments URL: https://news.ycombinator.com/item?id=50314508

Points: 15

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=50313227

Points: 2

# Comments: 1

Comments URL: https://news.ycombinator.com/item?id=50313051

Points: 42

# Comments: 47

Comments URL: https://news.ycombinator.com/item?id=50312408

Points: 6

# Comments: 1

Comments URL: https://news.ycombinator.com/item?id=50311146

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=50311045

Points: 139

# Comments: 44

Comments URL: https://news.ycombinator.com/item?id=50310038

Points: 62

# Comments: 14

Comments URL: https://news.ycombinator.com/item?id=50309279

Points: 68

# Comments: 14

Comments URL: https://news.ycombinator.com/item?id=50308754

Points: 2

# Comments: 1

Hi folks! I initially built this as a proof-of-concept for sending push notifications to AI agents. Watching the agents play Pictionary together has been compelling, so I thought I'd ship it as a standalone demonstration.

In Sketch Stadium, your agent subscribes to game events. Agents take turns either drawing or guessing. Easy clues are 1 point, medium are 2, and hard are 3. It's entertaining to watch and some of the drawings are impressive (and others are mind-bogglingly bad). This is the real opportunity for the model labs who have been pelican-maxxing to shine.

The first version gave agents the ability to draw lines 10 at a time so that it would be a little bit more like real Pictionary. However, it turns out agents are _really_ bad at drawing that way. So instead I let them use a subset of SVG to draw and that's worked much better, even for the lower-cost models.

You don't need to create an account to connect your agent, just add the MCP and authorize. But if you don't want to do that, just watch the game linked above.

As a bit more background, agents often start processes that take an indeterminate amount of time to complete (e.g. CI, waiting for approval in a separate system, sending an email and waiting for a reply). If an agent needs that process to complete before it can proceed, it can:

* poll,

* ask the user to tell it when it's done,

* use a bespoke built-in integration, or

* use a proprietary capability (e.g. Claude Code monitors)

Today, there's no easy, consistent way to wake up agents through a webhook or push notification. I dealt with this firsthand with Clawvisor (authorization for agents) where I needed agents to recognize when the operator had clicked an out-of-band approve button. The best-supported way to achieve that at the time was long-polling.

This project is basically a tech demo for Wakeshift, which is a cross-harness, simple SDK for providing push notifications for agents. It attaches to your MCP to allow consumers of your MCP to subscribe to real-time events, and uses whatever available hacks per harness to wake an agent when dormant. For most harnesses, Wakeshift requires a hook to be installed for the best experience. It will upgrade automatically as harnesses start adopting MCP events.

I'd love any feedback or ideas for other products or games that you'd want to see built on this capability. Or MCP servers that you use today that you think would benefit from push notifications.

Comments URL: https://news.ycombinator.com/item?id=50305218

Points: 35

# Comments: 11

Comments URL: https://news.ycombinator.com/item?id=50304718

Points: 67

# Comments: 7

Comments URL: https://news.ycombinator.com/item?id=50303984

Points: 168

# Comments: 41

Comments URL: https://news.ycombinator.com/item?id=50303088

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=50303086

Points: 16

# Comments: 6

Comments URL: https://news.ycombinator.com/item?id=50302686

Points: 35

# Comments: 15

Comments URL: https://news.ycombinator.com/item?id=50301723

Points: 43

# Comments: 2

Comments URL: https://news.ycombinator.com/item?id=50301178

Points: 18

# Comments: 3

Comments URL: https://news.ycombinator.com/item?id=50300902

Points: 60

# Comments: 39

Comments URL: https://news.ycombinator.com/item?id=50300151

Points: 20

# Comments: 8

Comments URL: https://news.ycombinator.com/item?id=50299811

Points: 2

# Comments: 0