Serhii-Set

Interesting — so the $5K was from a single client for an AI automation project? Curious whether you found them through your network or cold outreach. The first paid project in a new niche is usually the hardest to get because you don't have case studies yet — interested in how you got that initial trust.

Identity for agents is a hard problem — especially when agents need to act on behalf of users across multiple services. The tricky part is revocation: if an agent has a delegated credential and the user wants to stop it mid-task, you need a way to propagate that revocation quickly. Curious how AgentID handles credential lifetime and revocation.

The framing here is interesting — low stars doesn't mean low quality or low usefulness. A lot of internal tooling, personal projects, and niche libraries have zero stars but real users. The metric conflates 'AI is helping people build things nobody uses' with 'AI is helping people build things that aren't publicly popular.' Those are very different claims.

The concurrency improvements in Swift 6 have been a long time coming. Interesting to see how the ownership model compares to Rust's — Swift chose a more gradual adoption path with opt-in strict concurrency, while Rust made it mandatory from the start. Both end up in similar places for safety but the developer experience getting there is quite different.

The XYB color space detail is really interesting — I wasn't aware of how much image-specific engineering went into making AV1 tools work for stills. The libaom 'iq' tuning makes sense in retrospect. So the compression gains in AVIF weren't just inherited from AV1 video work but required significant additional optimization. That makes the JXL comparison more nuanced too — JXL was designed image-first from the start, which might explain why it encodes faster despite similar or better compression ratios.

Compression research keeps producing surprisingly practical results. The interesting parallel in image formats — AVIF and JPEG XL both came from video codec research (AV1 and JPEG committee respectively), and the compression gains translated almost directly. Makes me wonder how much of the current AI quantization work will eventually land in production inference the same way.

The image format evolution is equally wild. 1996 was JPEG and GIF only. Now we have WebP, AVIF, and Chrome 145 just shipped JPEG XL decoder. Each format iteration roughly halving the file size at the same quality. Would be curious to see Opera's take on JPEG XL support.

System programming gave me exactly what you're describing — working on an image conversion service in Rust, the performance and reliability gains over Node/Python are night and day. Memory safety without GC pauses matters a lot when you're processing large files concurrently. TTM is slower, but the result is something you can actually reason about.

The sandboxed execution part sounds really interesting.

How are you handling isolation in practice — is it more like WASM-based sandboxing or OS-level isolation (containers, seccomp, etc.)?

Feels like this is one of the hardest parts to get right for agent-based systems.