@sitnikcode

Author of @PostCSS. Currently trying to fix social media with https://xitter.cf/t.co/k9jnMullaW Russian: @andrey_sitnik 🦋 @en.sitnik.es 🐘 @sitnik_en@mastodon.social

 Barcelona, Spain
sitnik.es
Joined June 2017
  • Tweets17K
  • Following306
  • Followers7.4K
2.2KMedia
Include

Only show posts containing:

Exclude

Hide posts containing:

Time range
-
Minimum likes
.@browsercompany @joshm is sunsetting Arc and says Dia is the ideal replacement. I disagree. Arc is a labor of love and design craft, and I enjoy using it every day. Dia, on the other hand, lacks that design love, and I’m not sure why. Here’s 6+ things where Arc is still better than Dia
54
65
26
1,040
44,035
I just tested offline mode on my flight from Barcelona to Tbilisi. I'm speaking at a meetup there tomorrow about all the privacy and security safeguards we've built into Slow Reader. tbilisijs.dev/events/tbilisi…
1
135
Added offline support to Slow Reader, including preloading images, etc. Offline support is standard for RSS readers. But now you can read your social feeds offline, too even on the NYC subway (where signal is very bad) or a plane. slowreader.app/
1
1
1
286
Slow Reader now loads your Mastodon and Bluesky timeline. Other networks are coming later. Why the white screenshot instead of our usual warm yellow? The theme changes when you switch from Useful to Fun, gently nudging you to stay in Useful. Releasing today.
1
2
245
Nuclear Design Bureau has come a long way since it's initial release, now we have physics improvements, leaderboards, challenge missions and a whole host of physics visualizations modes for you simulations. I'm amused at how this can stress your GPU while looking like this:
5
27
2
453
22,370
I want to go even further: a browser extension or mobile app could verify that every connection reaches a server running the expected Docker image. There's a cool technology for that: hardware attestation, where a CPU can prove what's running. But that's for later.
1
2
107
The verifier itself is intentionally simple: no dependencies, with literally every line of code commented. You can audit it yourself (or ask an LLM to help). It's designed to keep running automatically, without my involvement.
1
1
72
Click any check for step-by-step instructions to reproduce it on your own machine. And when you open the verifier, its HTML checks in your browser that Slow Reader's JS and HTML hashes match the latest verified results.
1
1
15
The verifier checks that: 1. The site is online 2. A release exists for the deployed commit 3. Rebuilding gives identical hashes 4. GitHub CI and I both signed the image 5. The live JS and HTML match the release's manifest.json
1
1
34
But how can regular users verify all that? I made a separate CI repo that runs the checks every hour, publishes a public dashboard, and writes the results to an append-only log. hplush.gitlab.io/slowreader-…
1
1
21
Every image includes a manifest.json with the commit SHA and hashes of all JS and HTML files. The JS hashes also go into the import map's integrity field. If a JS file changes, the browser refuses to load it. slowreader.app/manifest.json
1
1
26
GitHub Actions builds and signs the image in CI, so you can verify its provenance. Then I sign the release with my own YubiKey, so nobody else can publish an authorized release. github.com/hplush/slowreader…
1
1
79
Our production servers run the exact same Docker image we publish for self-hosting. The build is reproducible: build it from the repo on your own machine, and you'll get the same image digest. github.com/hplush/slowreader…
1
1
118
Why? I want you to know that nobody can see what you read. The app downloads your feeds directly. Our server only handles sync, and that data is end-to-end encrypted. But I (or law enforcement) could still inject surveillance code into the app's JavaScript.
1
1
160
I want Slow Reader to set new standards for protecting users. It's open source, local-first, and end-to-end encrypted. But how can you prove a website was actually built from its published source code? Thread ↓ hplush.gitlab.io/slowreader-…
1
2
4
532
🚨 LiteLLM, a Python package for connecting to AI model providers with 92M+ monthly downloads, has been compromised. Version 1.107.0 ships credential-stealing malware through PyPI wheels and Docker images, targeting cloud credentials, developer secrets, and AI tools.
111
324
176
1,907
295,645
If you moved to Barcelona on a highly qualified professional visa, there's a city service that can help you navigate bureaucracy in English. I know someone who got a TIE appointment through them in just a couple of days. barcelona.cat/internationalw…
1
3
327
Replying to @happywebcoder
The current idea is to lets users to generate with LLM own plugins for cases like IIRC, etc
1
11
I'm building Slow Reader to read more useful threads, not rage bait. I liked Bluesky's cute thread indicator, so I added it to the UI. It'll work across all social networks, with more UX tweaks and gamification to help break bad scrolling habits. slowreader.app/
1
1
6
528
While people stuck in Musk’s algorithmic bubble think the EU has no AI, France’s Mistral just unveiled Large 4, competitive with top Chinese open-weight models. The EU also has strong OCR and video generation models, plus promising AI chip startups like axelera.ai/
2
5
413
Load more