Rogue

I like to yap about AI news. Im also a programmer.

[Rogue @Rogue0114]

I made this project, which started at the beginning of February, and I called it diffAdvisor. It is a Tauri app that uses AI to review your code, aimed at being a learning tool for students and juniors. It also builds a knowledge graph with MD files that can be opened in Obsidian.

The video below shows the full app, with simple mock examples of all the data.

I started the project about a week after Codex Desktop launched, and I finished it a day after Codex was released on Windows, at the beginning of March. While using Codex, I noticed I could just create a SKILL for Codex, together with some automations, tell it to create files in Markdown style, and my idea would be done just like that. So, while I postponed this post a lot, I wanted to at least share what I was building. I am already working on my next project, but it is very experimental, so I am not sure I will talk about it.

Now, about diffAdvisor:

The review is more high-level than just criticizing the files it reads line by line. It happens at a more architectural level, and it questions the user on their own knowledge of the code. It also creates .md files that connect different concepts it sees in the code, with the goal of growing a database of raw data in a more fundamental and generalized way.

It also gives ideas on different decisions the user could have made, like using a different library to achieve a certain result. The model knows what to check based on the system prompt it has, together with SKILLs activated based on the project it is reading, to look for things more specific to that language or framework.

All the files can be read and edited within the app interface, but the real goal was to open the text files with Obsidian as an incentive to keep learning and creating your own database, seeing how different concepts talk to each other, what their connections are, and using the texts for future reference.

And that’s about it. I made it while questioning how people should study programming in a world that now only generates AI code. The answer I found was that we need to focus even more on high-level decisions, knowing the “why” instead of the “how” behind the things we build.

[Rogue @Rogue0114]

@patience_cave You're gonna love the newest benchmark drop

https://t.co/quzy0qWHbl

[Programmer Humor @PR0GRAMMERHUM0R]

newBenchmarkDropped

[Rogue @Rogue0114]

Amazing. But at what cost? 😔

[ClaudeDevs @ClaudeDevs]

We've reset 5-hour and weekly rate limits for all users.

[Rogue @Rogue0114]

At least some piece of good news. Opensource is the only hope if this keeps up

[Bindu Reddy @bindureddy]

LOL! Forget Fable 5

Just got our benchmark results

Kimi 2.7 code is MIND BLOWING when it comes to agentic coding

Will publish shortly!

[Artificial Analysis @ArtificialAnlys]

Today is the first time our Intelligence Frontier chart has moved backward.

[Rogue @Rogue0114]

@scaling01 I used to think Trump was the type of crazy guy to let AI run wild no matter how bad it could get. I think i was wrong about him many times. Always proving to be worse than i thought

[Rogue @Rogue0114]

@Angaisb_ Man, i just switched jobs and was so excited to ship code with Fable. Our only hope now is gpt 5.6

[Rogue @Rogue0114]

@theo At least we don't have to think of a $1000 subscription i guess 😔

[Rogue @Rogue0114]

@doug_wolff @axiosalex @m_ccuri Reality was worse than your prediction

[Rogue @Rogue0114]

@elder_plinius With this face you look like the reason the gov did it 😂

[Lisan al Gaib @scaling01]

The US government just killed billions in Anthropic revenue over a non-universal jailbreak that exposes capabilities that other models like GPT-5.5 also possess

[Rogue @Rogue0114]

@Angaisb_ Yeahhh. I'm refreshing the page just seeing everyone i follow losing their minds. I just wanted to comment from the future. I'm still shocked

[Rogue @Rogue0114]

@AnthropicAI No way

[Chris @ChrissGPT]

Claude 5 Fable (Ultracode)

I asked it to build a demo of my dream game in Three.js and I'm genuinely shocked 💀

One shot, a full explorable starship with a working cockpit, crew quarters, a planet drifting past real windows, dynamic lighting, sleep/eat interactions,

it screenshotted its own work and fixed itself until it hit 60fps on browser

Obviously not steam ready but man this is so so far from what we had one year ago…

[Rogue @Rogue0114]

@bindureddy i hope thats all true

[Rogue @Rogue0114]

@Angaisb_ I didn't see rumors of them lowering the API costs. If this is true then its massive.

[Rogue @Rogue0114]

@uzairakrum @VictorTaelin in theory, yes. but i guess he wants to be as fast a possible and with a project this big of this much complexity it would be very hard to hire someone for it, make them understand the current code and also receive more than a months payment. this isnt a juniors project either.

[Rogue @Rogue0114]

@VictorTaelin I have to say this is a strong indicative of the cybersecurity capabilities Anthropic claimed the model has. Im glad it is that good, at least for you.

[Chubby♨️ @kimmonismus]

Claude 5 Fable tl;dr

- It is state-of-the-art on nearly all tested benchmarks of AI capability, showing exceptional performance in software engineering, knowledge work, vision, scientific research

-The longer and more complex the task, the larger Fable 5’s lead over our other models

-its more token-efficient than past Claude models

- Fable 5 stays focused across millions of tokens in long-running tasks and improves its outputs using its own notes

Fable 5 is more than just better benchmarks. It's more efficient, allows for longer work periods, offers better context management, and so much more.

GPT-5.6 is just around the corner.

I'm a huge Codex fan, but Fable/Mythos is in a league of its own. I'm curious to see if OpenAI will release its own Mythos.

"During early testing, Stripe reported that Fable 5 compressed months of engineering into days. In a 50-million-line Ruby codebase, the model performed a codebase-wide migration in a day that would otherwise have taken a whole team over two months by hand."

[Rogue @Rogue0114]

@scaling01 depends. whos gonna use the model with api only?