Skip to content
AI NEWS JOURNAL
AI, IN PERSPECTIVE

Search AI news

Search headlines, summaries and articles to connect the dots.

Newest first

Show HN: Gitfold – every Git branch as a folder, ready in a second

Hi HN!I want to show you a project which I'm using myself to save time on worktrees creation and management. My case is a bit specific because I'm normally working on a really big monorepo, where creation of a worktree takes around a minute and 20+GB of disk space. Previously I was using a pool of worktrees which were managed by my own sessions manager. It worked, but was a bit clunky and sometimes I was running out of the worktrees anyway.As a fun side-project I decided to create a VFS for serving git branches directly from the git object store, with only one real checkout needed. I started with read only version to speed-up the reviews (I often checkout the PRs code to get LLM support) and that is very easy to do with FUSE, but it would not let me fully replace my old solution. Adding write capabilities was a bit more involved, and writing without full kernel passthrough turned out to be a few times slower. But interestingly for my workflows it does not matter. I was also running some external tests like full Linux kernel build (allmodconfig) and the performance hit vs a native build was around 2.5%. So far so good!After building the core I realised that having VFS allows me to play with other interesting features. And I implemented a few:- tree structure of branch dependencies; you can e.g. have branch 'api', on which 'store' and 'ui' depend. When you add or fetch new commits to 'api', whole tree rebases automatically- read only views showing "merged" state of each node in the tree from the previous point; e.g. you can check how 'api' branch would look like, if both 'store' and 'ui' would be merged as is; currently it has two modes: only committed or all changes- command for showing conflicts between current worktree and all other open worktrees- command for showing all worktrees touching given file of interest- agents getting real time notifications about potential conflictsWhat is also important is that it works without an issue with all native git commands. Git sees those workspaces as normal worktrees. So you can e.g. normally rebase every such worktree separately (which is one of the main differences compared to overlayfs). I'm sure some people will think this project is very similar to GitFS, EdenFS, OverlayFS, ArtifactFS, etc... spoiler alert: it is not. Most of those projects are focusing on fetching objects from remote git server lazily. Which I can imagine is very useful if you need a fresh checkout on a new instance (eg. in a CI). But that is not my use case. I have a full local checkout, and need to be able to create and delete on demand a lot of worktrees. That is it :) But I also prepared a full comparison, just in case: https://github.com/pkukielka/gitfold/blob/main/docs/comparis...I'm happy to answer any questions! Comments URL: https://news.ycombinator.com/item?id=50601881 Points: 1 # Comments: 0

Perplexity and Intelligence

How perplexity measures progress in AI models, and why it could help developers establish a shared approach to evaluating emerging capabilities.

Show HN: A modern take on the Go language

Targets Go to use the toolchain and the package system you already know and love Comments URL: https://news.ycombinator.com/item?id=50595828 Points: 2 # Comments: 0

Show HN: Hikari.ImGui, a pure C# re-implementation of Dear ImGui

A pure, standalone C# re-implementation of the much-beloved and widely used immediate-mode GUI for 3D applications, Dear ImGui, targeting version v1.93.0-docking with zero native dependencies. Please feel free to use for your projects. Also available as a nuget package. Comments URL: https://news.ycombinator.com/item?id=50595885 Points: 2 # Comments: 0

Show HN: Reviving a 2012 open-source password manager for modern Android

Offline password manager for Android: a fork of Passdroid (last release 2012), brought up to date. The app has no internet permission.Two key new features: - 2FA codes (TOTP) - fingerprint unlockPlus many smaller changes and fixes, and security hardening: AES-256 with PBKDF2 at 600k iterations, AES-GCM encrypted exports. Threat model and design are in INTERNALS.md.Built largely with Claude Code; the design and security decisions are mine.Feedback welcome, especially on the security side. Comments URL: https://news.ycombinator.com/item?id=50594645 Points: 1 # Comments: 0

Sovereign AI: Where control becomes an architectural requirement

We believe sovereignty is becoming an architectural requirement for firms operating globally, especially in highly regulated industries. The business question for buyers is how much control an organization needs over the systems that produce its intelligence, and what happens when that control is ceded to others. U.S. companies supply most of the cloud and frontier […] The post Sovereign AI: Where control becomes an architectural requirement appeared first on SiliconANGLE.

Sovereignty starts with data, not chips, says Stuttgart AI factory chief

A sovereign AI factory is only as valuable as the companies that can actually get onto it. Across Europe, publicly funded AI factories are racing to give startups and small manufacturers supercomputer-class GPUs without pushing them into lock-in or loose data handling. In Stuttgart, the HammerHAI AI factory is taking shape around a system from […] The post Sovereignty starts with data, not chips, says Stuttgart AI factory chief appeared first on SiliconANGLE.

AI deals shrink to as little as 30 days as the data layer sets the pace

Artificial intelligence has turned data infrastructure from a back-office purchase into the bottleneck that decides how fast GPUs pay off. As hyperscalers, neoclouds and enterprises race to expand their data layers, the way that infrastructure gets bought is changing just as fast. The storage industry’s longtime performance specialist sees that shift as its opening. DataDirect […] The post AI deals shrink to as little as 30 days as the data layer sets the pace appeared first on SiliconANGLE.

Cut self-hosted LLM cost and latency with the distributed KV caching agent skill

Self-hosting LLMs on GPUs means paying to recompute the same context on every turn. Distributed KV caching with Amazon ElastiCache can cut inference cost by up to 40% and time to first token by up to 80%. This post walks through the distributed KV caching skill in the Agent Toolkit for AWS, which evaluates and sets up the cache from your coding agent.