Stanislav's Avatar

Stanislav

@khromov.se

πŸ‡ΈπŸ‡ͺ Web developer and obscure trivia enjoyer 🌱 Find me at https://stanislav.garden/ πŸ”΅ Starting over on Bluesky

920
Followers
202
Following
226
Posts
10.06.2023
Joined
Posts Following

Latest posts by Stanislav @khromov.se

🚨🚨🚨

PSA: if you were using the Svelte Claude marketplace: we've renamed the repo which means that unfortunately you would not get updates if you added the marketplace as `sveltejs/mcp`.

Please remove the old marketplace and re-install `sveltejs/ai-tools`...it comes with a surprise πŸ‘€

🚨🚨🚨

27.02.2026 21:28 πŸ‘ 19 πŸ” 8 πŸ’¬ 2 πŸ“Œ 1

Thanks!

27.02.2026 14:05 πŸ‘ 0 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0

πŸ‘‹ I'd like to to LocalFirst but not sure I can make it. Are the tickets refundable and/or transferable?

27.02.2026 13:16 πŸ‘ 0 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0
Preview
Overview β€’ Svelte MCP Docs Overview β€’ Svelte MCP documentation

We released a new version of the svelte opencode plugin. It now ships with all the available skills and automatically adds them to your opencode skills (require last version of both)

If you want you can disable the MCP or the Skill by creating a `svelte.json` next to your `opencode.json` config 🧑

30.01.2026 17:30 πŸ‘ 25 πŸ” 4 πŸ’¬ 1 πŸ“Œ 1

πŸ™‡ Let's make AI write Svelte even better! File an issue for tasks where LLMs are struggling to help us fix them!

13.01.2026 19:17 πŸ‘ 8 πŸ” 4 πŸ’¬ 0 πŸ“Œ 0
Preview
Join Claude! | Claude You've been invited to try Claude

If you haven't tried Claude / Claude Code yet, here's a few referral codes to get a free week. Have fun and build something awesome!

claude.ai/referral/3eB...

10.01.2026 00:42 πŸ‘ 2 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0

the anthropic drama (tl;dr they blocked third party tools from using claude code subscriptions) is a useful reminder that you use LLMs at the pleasure of the companies providing them

sounds obvious but despite all the AI discourse, i don't see a lot of people talking about the consequences of this

09.01.2026 23:51 πŸ‘ 280 πŸ” 35 πŸ’¬ 5 πŸ“Œ 10
Preview
Overview β€’ Svelte MCP Docs Overview β€’ Svelte MCP documentation

We released a Svelte Skill! It will teach your agent to use the svelte CLI to invoke the same tools as the MCP! We also released a subagent so that docs and iterations with the autofixer don't take up precious context!

And you can get all of that easily because the repo is now a Marketplace! 🧑

04.01.2026 22:08 πŸ‘ 44 πŸ” 12 πŸ’¬ 1 πŸ“Œ 0
Video thumbnail

Finally, opencode supports MCP prompts (this was the PR I was waiting to get released)! I already switched to it as my daily driver, now it will be much better (prompts are a powerful MCP primitive and I use it every time with the svelte MCP).

Next: resources! I already have them working locally! 😎

31.12.2025 18:44 πŸ‘ 17 πŸ” 1 πŸ’¬ 1 πŸ“Œ 0
Post image

Thanks for trying it out! I think the cost calculations weren't working right but after merging the below PR they should work (more or less) correctly. The main challenge is doing things in a provider-agnostic way.

github.com/sveltejs/ai/...

29.12.2025 17:34 πŸ‘ 3 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0

Works anywhere where you can use the Pro/Max plans!

28.12.2025 10:32 πŸ‘ 1 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0
Post image

πŸ§‘πŸ’» If you're using Claude Code on a Mac, the Claude Usage app lets you easily keep track of your usage from the menu bar, and it can even notify you when you're getting close to your limit. Free and open source at github.com/hamed-elfayo...

28.12.2025 09:45 πŸ‘ 21 πŸ” 1 πŸ’¬ 3 πŸ“Œ 0
Post image Post image

Work on SvelteBench 2.0 is in progress! This time we're working on making it fully agentic and to support the official Svelte MCP! If you have ideas for the next version, please share them at github.com/sveltejs/ai/... - some early screenshots below.

27.12.2025 09:45 πŸ‘ 26 πŸ” 2 πŸ’¬ 2 πŸ“Œ 1
This AI model writes Svelte 5 code PERFECTLY
This AI model writes Svelte 5 code PERFECTLY 00:00:00 Why AI struggles with Svelte 5 00:01:02 SvelteBench: early results 00:01:51 How scores improved over time 00:02:16 Open weights models 00:03:02 Small models that you can run locally 00:03:26…

πŸ‘€ The first model that writes perfect Svelte 5 code is here! In the video below you'll find out which one and see the scores for frontier models among all the big providers like Google, OpenAI and Anthropic! #svelte #javascript #ai

www.youtube.com/watch?v=nXIM...

26.12.2025 09:45 πŸ‘ 11 πŸ” 2 πŸ’¬ 0 πŸ“Œ 0
Join the Svelte Society Sweden Discord Server! Check out the Svelte Society Sweden community on Discord - hang out with 204 other members and enjoy free voice and text chat.

Welcome Moritz! Join the Swedish Discord at discord.gg/R2WWaYHG8W and Society Swe group at guild.host/svelte-socie...

We have a Julebord on the 15th Dec in case you want to quickly meet some people here, details in Discord. cc @kevinak.se

12.12.2025 15:16 πŸ‘ 3 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0
Preview
Custom Renderers for Svelte with Paolo Ricciuti In this episode of Svelte Radio, we welcome Jeppe as a new permanent...

πŸŽ™οΈ Just published a new episode of Svelte Radio: Custom Renderers for Svelte with Paolo Ricciuti . Have a listen:

11.12.2025 16:00 πŸ‘ 9 πŸ” 2 πŸ’¬ 0 πŸ“Œ 3

Wow, the future is here πŸ˜‚

12.12.2025 13:02 πŸ‘ 0 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0

The benchmark is open source Rodrigo, have a look! :-)

12.12.2025 12:16 πŸ‘ 1 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0
Post image

πŸ‘€ The first AI model from OpenAI that can properly write Svelte 5 is here - welcome GPT 5.2! All results at khromov.github.io/svelte-bench...

12.12.2025 11:08 πŸ‘ 18 πŸ” 1 πŸ’¬ 1 πŸ“Œ 0
Post image

πŸ’« Putting the finishing touches on my next video - tomorrow we'll find out how AI models have been improving at writing Svelte 5 code and which model has gotten a perfect store on the Svelte LLM benchmark! #ai #svelte

07.12.2025 13:58 πŸ‘ 14 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0
Post image

Anthropic enabled Opus 4.5 for all Claude Code users (including Pro!) and you can set /model opusplan to use Opus 4.5 only for Plan mode and Sonnet 4.5 for execution to make your quota last longer!
#claude #ai

03.12.2025 23:06 πŸ‘ 7 πŸ” 1 πŸ’¬ 0 πŸ“Œ 0
Preview
Security Advisory: CVE-2025-66478 A critical vulnerability (CVE-2025-66478) has been identified in the React Server Components protocol. Users should upgrade to patched versions immediately.

If you are running projects on Next.js, you probably want to immediately upgrade them since a vulnerability with CVSS 10.0 (highest possible) was just disclosed in the RSC implementation, meaning the door is essentially wide open to hack your sites.
nextjs.org/blog/CVE-202...

03.12.2025 21:59 πŸ‘ 1 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0
Post image

Guys I'm not making this up...I asked Claude, ChatGPT, Gemini and Grok what frontend framework they would pick if they only could pick the one they would enjoy writing more.

They ALL said Svelte 🧑

Do your LLM a favor, switch to Svelte 🧑

02.12.2025 11:44 πŸ‘ 21 πŸ” 4 πŸ’¬ 3 πŸ“Œ 0
Post image

A historic first today as Opus 4.5 scores 100% on SvelteBench - the first model to do so. We're hard at work on the next version of SvelteBench that will use an agentic approach, if you have feedback, submit it here: github.com/khromov/svel... #svelte #ai

25.11.2025 09:45 πŸ‘ 70 πŸ” 10 πŸ’¬ 3 πŸ“Œ 0

pretty excited about this feature (thanks @ottomated.net!). dealing with file uploads can be a bit of a pain β€” the way it works normally is that you (or your framework) does `await request.formData()` to get a FormData object. but it has to buffer every file into memory first. that's a problem!

21.11.2025 13:51 πŸ‘ 83 πŸ” 8 πŸ’¬ 2 πŸ“Œ 0
Post image

πŸ€– Gemini 3.0 SvelteBench results are out! Gemini 2.5 was already great and 3.0 improves on the already excellent scores. All results at khromov.github.io/svelte-bench... #svelte #ai

19.11.2025 09:45 πŸ‘ 15 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0
Post image

My quota limit for quota limits has been exceeded. πŸ™ˆ Just #AWS things!

12.11.2025 14:05 πŸ‘ 2 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0
Benchmarking AI with Stanislav Khromov
Benchmarking AI with Stanislav Khromov In this episode, Stanislav Khromov joins the Svelte Radio team to discuss his work on Svelte Bench, a benchmarking tool that scientifically measures how well different LLMs understand and write…

🎧 A couple of weeks ago I was a guest on the Svelte Radio podcast to talk about Svelte AI benchmarking and the brand new Svelte MCP server. Check it out in case you missed it! #svelte #javascript
www.youtube.com/watch?v=6PTJ...

10.11.2025 09:45 πŸ‘ 11 πŸ” 3 πŸ’¬ 0 πŸ“Œ 0

If you're not tired of hearing about Svelte and AI I'm was recently on the Svelte Radio podcast!

30.10.2025 18:50 πŸ‘ 11 πŸ” 3 πŸ’¬ 0 πŸ“Œ 0
Video thumbnail

Interested in hearing about Svelte and AI? Tomorrow we're releasing an episode of Svelte Radio with @khromov.se to talk about Svelte Bench and the new Svelte MCP server that was recently released.

We also talk about how LLMs always want to write React

Stay tuned!

29.10.2025 18:59 πŸ‘ 16 πŸ” 4 πŸ’¬ 0 πŸ“Œ 0