加密内参

Why do more and more people in the city love raising Chinese farm dogs?

CAD engineers can’t sleep after seeing this

I watched the demo of text-to-CAD, an open-source project that generates a complete 7-DOF mechanical arm from natural language, complete with kinematics, a custom GUI, and then exports to STEP, slices it, and sends it directly to a printer.

I used to think this was just something to mess around with. I didn’t expect it could actually output usable STEP files.

Let me share a real feeling: the most torturous part of building a hardware prototype isn’t drawing the diagrams itself—it’s the wall between CAD and code.

After finishing the model, you have to manually export URDF, write coordinate transforms, tweak a bunch of parameters, and a single person ends up juggling six or seven tools. One wrong bump in the pipeline and it crashes. The intermediate format “agent” produces can’t be read at all.

This harness turns every step—modeling, exporting, describing, slicing, and sending—into functions that an agent can call.

With Claude Code and Codex as the orchestrators, you only say one line: “I want a mechanical arm that can grab things,” and it handles the rest.

The generated geometry is probably rough for now, but don’t forget its iteration speed.

Previously, changing one joint position meant: open SolidWorks, edit the sketch, update the assembly, re-export the URDF, and re-compute kinematics. Now you change the prompt and, ten seconds later, you can see a new model. That broken pipeline gets killed by a file.

What makes it feel even more worth it is that it runs locally. It doesn’t depend on the cloud, and iterating through trial and error doesn’t come with API-cost anxiety. Developers should be willing to pay for this kind of freedom.

Unclear part: with complex assemblies that include gears, tolerances, and cables, can it really be done well with pure prompts?

But the starting point is already far lower than I imagined.

#AI硬件 #开源工具 #CAD

RT @thebuggeddev: My honest take on @Kimi_Moonshot K3 👀

As soon as I got access to Kimi, I put it through the exact same 3D football stadi…

The Most Overrated Modern Chinese Writers

Every morning, the moment I open my eyes, I open Twitter to see if the account is still there

Codex desktop app: 10,000 skin swaps in just 4 days.

Codex Dream Skin—an instant skinning tool that doesn’t modify the official package, doesn’t touch .asar, and uses pure local CDP injection. It reached the top of GitHub Trending in less than a week after release.

Its mechanism is so simple it can be explained in one sentence: a 16:9 wallpaper + CSS injection completely reshapes the atmosphere of the entire Codex home page.

And that’s exactly what makes it interesting.

Codex is a desktop IDE built by OpenAI for AI programming. The need to skin it has existed in the traditional IDE ecosystem (VS Code / JetBrains) for a decade. But when an AI-native programming tool appears, the first community-level request from users isn’t to add features—it’s to put on a good-looking face.

Something is happening: AI programming tools have entered the next phase.

Once feature competition converges, what’s left to fight for is the tool’s emotional value.

This follows the same logic as when, after jailbreaking the iPhone, everyone first installed WinterBoard to change themes. When a tool reaches a certain level of maturity, users’ attention shifts to the tool’s appearance.

If you’re using Codex too, it’s worth spending 5 minutes installing it.

4:6 Is this an official match?

It’s expected that within these next two days,
Deepseek V4 (GA) will be released.
According to feedback from users who were included in the rollout tests,
its coding ability is on par with Opus 4.8,
but the price is an order of magnitude cheaper.

This will deal a devastating blow to the valuations of Claude and OpenAI.

RT @SunNeverSetsX:

RT @li9292: Kimi K3 today topped the Arena programming leaderboard.
In the same company, they’re dismantling their own corporate structure
because the HKEX doesn’t recognize its current structure.

In the Dark Side of the Moon, valuation $4.3B→$31.5B, only took 6 months.

I asked Apodex @Apodex_AI one question:
Is this company worth that price…

Life Show of African Brothers

On the timeline of “pushing,” how long has it been since you last saw Gemini 3.1, Nano Banana 2, or VEO3 information?
It seems the big model team at Google has collectively hit the pause button.

Where did that all-powerful Google go?

It’s said this is the girlfriend of Spain’s teenage prodigy, Yamal. After Spain beat France 2-0, the girlfriend also celebrated the victory with two goals.

If you didn’t receive a Kimi K3 merchant order today, it means you’re still an undistinguished AI blogger.

Argentina staged a comeback against England to reach the final—do fans who only see the game every four years think something is off about this match?

Today's French supercar group is really desperate. They met their nemesis Spain and missed out on the final.

After this match ended,

the Norwegian player number 7 said:
"I originally planned to pass to Haaland,
but at the time it felt like there wasn't any space to make the pass."

Brothers, do you think he should have passed that ball?

Why is the Argentina team mocked by the whole internet?
Mainly because there are too many football fans with the right vibes every four years

After watching Stephen Chow's current movies, I regret it for ten years; if I don't watch them, I regret it for one year

A good skill is used by others → creates value → feedback helps improve it → the skill gets better → more people use it → exponential spread and iteration.

This is exactly the same as past software open-sourcing: Linux, Python, React—create it once, reuse it globally, and keep evolving.

The only thing you need to judge is whether this skill can stand on its own as an IP—whether it’s high-value, has a broad audience, can be iterated continuously, and is easy to spread.