Hacker News

Article URL: https://www.draketo.de/software/programs-you-can-still-hack-when-dumb.html

Comments URL: https://news.ycombinator.com/item?id=48077712

Points: 1

# Comments: 0

Article URL: https://ai-app-store-screenshots.vercel.app/

Comments URL: https://news.ycombinator.com/item?id=48077702

Points: 1

# Comments: 0

Article URL: https://github.com/ammaarreshi/gemma-chat

Comments URL: https://news.ycombinator.com/item?id=48077698

Points: 1

# Comments: 0

Article URL: https://www.npr.org/2026/05/06/nx-s1-5810941/new-orleans-ocean-study

Comments URL: https://news.ycombinator.com/item?id=48077689

Points: 1

# Comments: 0

Article URL: https://www.slowboring.com/p/why-movies-are-getting-longer

Comments URL: https://news.ycombinator.com/item?id=48077634

Points: 2

# Comments: 0

Article URL: https://www.sciencedirect.com/science/article/pii/S1383571821000644

Comments URL: https://news.ycombinator.com/item?id=48077628

Points: 1

# Comments: 0

I had been looking to try <500M parameter language models but you wouldn't find an API to try them anywhere, so I built this cloudflare hosted static website that hosts weights and built an inference runtime for these models that uses WebGPU and runs inference from your browser.

These are only so useful in a multi-turn conversation but it's still interesting to see what you can pack in a <250mb model.

I tried using ONNX versions earlier, but there were too many quirks of using them with language models and the TPS wasn't too impressive. Inspired by svenflow/webgpu-gemma, I put my codex and claude to the task of writing WGSL to run inference for GGUF versions of these models.

Once you load this website and a model, it should load offline too, until your browser evicts the model from the cache.

Comments URL: https://news.ycombinator.com/item?id=48077627

Points: 3

# Comments: 0

Article URL: https://twitter.com/talhof8/status/2051337721151455509

Comments URL: https://news.ycombinator.com/item?id=48077626

Points: 1

# Comments: 0

Article URL: https://www.astralcodexten.com/p/three-model-organisms-for-taste

Comments URL: https://news.ycombinator.com/item?id=48077625

Points: 1

# Comments: 0

Article URL: https://blog.noforeignland.com/off-grid-boat-communications-with-meshtastic/

Comments URL: https://news.ycombinator.com/item?id=48077624

Points: 1

# Comments: 0

Article URL: https://chrismorgan.info/no-query-strings

Comments URL: https://news.ycombinator.com/item?id=48077608

Points: 2

# Comments: 0

Article URL: https://taskandpurpose.com/tech-tactics/army-xm30-bradley-replacement/

Comments URL: https://news.ycombinator.com/item?id=48077602

Points: 1

# Comments: 0

Article URL: https://midnightcheese.com/2026/05/aprs-message-36-miles-two-ht-radios/

Comments URL: https://news.ycombinator.com/item?id=48077590

Points: 1

# Comments: 0

Article URL: https://enufstyle.com/generators/bart/

Comments URL: https://news.ycombinator.com/item?id=48077529

Points: 3

# Comments: 0

Article URL: https://articles.adsabs.harvard.edu/pdf/1980QJRAS..21..267T

Comments URL: https://news.ycombinator.com/item?id=48077527

Points: 2

# Comments: 1

Article URL: https://github.com/JosefAlbers/mlx-code

Comments URL: https://news.ycombinator.com/item?id=48077522

Points: 1

# Comments: 0

I'm Anders, co-founder of Shoehorn (https://shoehorn.dev/).

For about a year I was genuinely excited about Backstage. Saw the Spotify demo, thought "I want that too". Turned out it wasn't as easy. But to be fair, Backstage is a great piece of software. Powerful, flexible, and if you have a dedicated developer team to run it, you can build something amazing on top of it.

The catch is exactly that: you need a dedicated team. Most companies don't have one. Or they do, and that team would rather be solving product problems than maintaining the platform that helps everyone else solve theirs. We wanted something that gave teams 90% of the value with 10% of the operational overhead.

We built Shoehorn out of frustration.

Goals from the start:

1. A great UI, API first, Terraform and CLI: That's what most users actually want.

2. Surface problems clearly: No silent failures, no "why didn't my service show up." If something's wrong, you see it.

3. Automate the boring stuff: The data is already in GitHub, Kubernetes, your identity provider. Use it. Don't make humans retype it into manifests.

4. Maintainable by one person: Deploy and use it. No team required.

Shoehorn is an event-driven developer platform — service catalog, scaffolding, scorecards, fuzzy search, multi-tenant by default. Same category as Backstage, Port, and Cortex. The difference: Backstage is a framework you assemble into a portal. Shoehorn is a platform you turn on. Auto-discovery from Kubernetes, GitHub, Cloud vendors etc.

Auto-discovery from Kubernetes and GitHub crawler for markdowns and (manifests if desired). One install, sensible defaults (we think), prebuilt hardened images from (https://docs.docker.com/dhi/), gives you an working catalog inside within 15 min.

Architecturally its a small set of Go microservices, its backed by Postgres(RLS), Meilisearch, Valkey and Redpanda. Authentication is OIDC, we currently support: Zitadel and Okta.

We released Beta yesterday, there's probably bugs and some parts might be rough. Would love to hear any thoughts, comments, or ideas of how we could make this better, and if there are obvious things we’re missing etc. Are you using backstage or any other internal developer portal today?

Demo page here: https://demo.shoehorn.dev

Comments URL: https://news.ycombinator.com/item?id=48077324

Points: 1

# Comments: 0

Article URL: https://isthisanart.substack.com/p/death-of-a-software-developer

Comments URL: https://news.ycombinator.com/item?id=48077317

Points: 1

# Comments: 0

Article URL: https://www.copahost.com/blog/dirty-frag-cve-2026-43284/

Comments URL: https://news.ycombinator.com/item?id=48077314

Points: 1

# Comments: 0

Article URL: https://x86.lol/generic/2022/11/26/lanzaboote.html

Comments URL: https://news.ycombinator.com/item?id=48077291

Points: 1

# Comments: 1