Reposted by ratevyra
We're using WeatherNext technology to upgrade weather forecasts in Search, Gemini, Pixel Weather and Google Maps Platform’s Weather API. In the coming weeks, it will also help power weather information in Google Maps. Learn more → goo.gle/4oO2gJY
WeatherNext 2: Our most advanced weather forecasting model
The new AI model delivers more efficient, more accurate and higher-resolution global weather predictions.
goo.gle
November 17, 2025 at 5:16 PM
We're using WeatherNext technology to upgrade weather forecasts in Search, Gemini, Pixel Weather and Google Maps Platform’s Weather API. In the coming weeks, it will also help power weather information in Google Maps. Learn more → goo.gle/4oO2gJY
Reposted by ratevyra
Added 2 New Primitives : Svelte Prompt Kit
1. Simple Chat App
2. Tool Calling with Vercel AI SDK + OpenRouter
Tip: If you hit rate limits, switch to an OpenRouter key (I’m testing with the FREE z-ai/glm model).
Live : svelte-ai-elements.vercel.app/prompt-kit/p...
1. Simple Chat App
2. Tool Calling with Vercel AI SDK + OpenRouter
Tip: If you hit rate limits, switch to an OpenRouter key (I’m testing with the FREE z-ai/glm model).
Live : svelte-ai-elements.vercel.app/prompt-kit/p...
November 16, 2025 at 10:56 PM
Added 2 New Primitives : Svelte Prompt Kit
1. Simple Chat App
2. Tool Calling with Vercel AI SDK + OpenRouter
Tip: If you hit rate limits, switch to an OpenRouter key (I’m testing with the FREE z-ai/glm model).
Live : svelte-ai-elements.vercel.app/prompt-kit/p...
1. Simple Chat App
2. Tool Calling with Vercel AI SDK + OpenRouter
Tip: If you hit rate limits, switch to an OpenRouter key (I’m testing with the FREE z-ai/glm model).
Live : svelte-ai-elements.vercel.app/prompt-kit/p...
Reposted by ratevyra
the ironic part about immigrants is they’re not lazy, the lazy ones didn’t have enough agency to move to a different country
immigration is as close to a filter for high performing individuals as you’re going to get
immigration is as close to a filter for high performing individuals as you’re going to get
November 16, 2025 at 6:35 PM
the ironic part about immigrants is they’re not lazy, the lazy ones didn’t have enough agency to move to a different country
immigration is as close to a filter for high performing individuals as you’re going to get
immigration is as close to a filter for high performing individuals as you’re going to get
Reposted by ratevyra
9 years ago from today @rich-harris.dev made the first commit to @svelte.dev 🍰
github.com/sveltejs/sve...
github.com/sveltejs/sve...
initial commit · sveltejs/svelte@fc7e6e6
github.com
November 16, 2025 at 10:31 AM
9 years ago from today @rich-harris.dev made the first commit to @svelte.dev 🍰
github.com/sveltejs/sve...
github.com/sveltejs/sve...
Reposted by ratevyra
K2 is hard to run in prod
A long list of inference providers have tried to run K2 but found the benchmarks much worse than the official API. Similar for gpt-oss
ML Ops is in fact not easy at all, and despite detailed instructions how to run it, nobody does it as well as Moonshot
A long list of inference providers have tried to run K2 but found the benchmarks much worse than the official API. Similar for gpt-oss
ML Ops is in fact not easy at all, and despite detailed instructions how to run it, nobody does it as well as Moonshot
November 16, 2025 at 12:25 PM
K2 is hard to run in prod
A long list of inference providers have tried to run K2 but found the benchmarks much worse than the official API. Similar for gpt-oss
ML Ops is in fact not easy at all, and despite detailed instructions how to run it, nobody does it as well as Moonshot
A long list of inference providers have tried to run K2 but found the benchmarks much worse than the official API. Similar for gpt-oss
ML Ops is in fact not easy at all, and despite detailed instructions how to run it, nobody does it as well as Moonshot
Reposted by ratevyra
after sleeping on it — K2 is definitely a viable model, Rust works fine but with caveats
accuracy — these models do fine pass@8, and claude code will basically automate those retries for you, which means accuracy isn’t a problem but:
cost — it’s expensive. even on a cheap model, Rust costs
accuracy — these models do fine pass@8, and claude code will basically automate those retries for you, which means accuracy isn’t a problem but:
cost — it’s expensive. even on a cheap model, Rust costs
impressions so far, ~1 hour in
* K2 in Claude Code is slow, and doesn't have a dramatically different feel from other models
* Rust with AI seems fine, mostly. There were a couple issues, but a CLAUDE.md fixed them
* K2 in Claude Code is slow, and doesn't have a dramatically different feel from other models
* Rust with AI seems fine, mostly. There were a couple issues, but a CLAUDE.md fixed them
lettuce begin
November 16, 2025 at 1:16 PM
after sleeping on it — K2 is definitely a viable model, Rust works fine but with caveats
accuracy — these models do fine pass@8, and claude code will basically automate those retries for you, which means accuracy isn’t a problem but:
cost — it’s expensive. even on a cheap model, Rust costs
accuracy — these models do fine pass@8, and claude code will basically automate those retries for you, which means accuracy isn’t a problem but:
cost — it’s expensive. even on a cheap model, Rust costs
Interesting results, although I also had Perplexity as the winner. I wasn't expecting OpenAI in second though...
At the AI Conference, attendees were asked: which startup would you short? What’s your pick in the evolving AI landscape#AIInnovationsns
https://www.businessinsider.com/at-an-ai-conference-attendees-were-asked-which-startup-they-would-short-2025-11
https://www.businessinsider.com/at-an-ai-conference-attendees-were-asked-which-startup-they-would-short-2025-11
At a top AI conference, 300 attendees were asked which startup they would short.
300 founders and investors in the audience at Cerebral Valley were asked which startup valued at more than a billion dollars they would bet against.
www.businessinsider.com
November 16, 2025 at 6:50 AM
Interesting results, although I also had Perplexity as the winner. I wasn't expecting OpenAI in second though...
Reposted by ratevyra
New SWE-Bench Verified results for GPT-5.1-Codex
it beats Sonnet-4.5 but is 26x cheaper (although 2.4x slower)
Surprisingly, GPT-5.1 is one of the cheapest models
it beats Sonnet-4.5 but is 26x cheaper (although 2.4x slower)
Surprisingly, GPT-5.1 is one of the cheapest models
November 15, 2025 at 10:19 AM
New SWE-Bench Verified results for GPT-5.1-Codex
it beats Sonnet-4.5 but is 26x cheaper (although 2.4x slower)
Surprisingly, GPT-5.1 is one of the cheapest models
it beats Sonnet-4.5 but is 26x cheaper (although 2.4x slower)
Surprisingly, GPT-5.1 is one of the cheapest models
Reposted by ratevyra
We are finally making a svelte skill for claude (and turning the repo into a claude marketplace so you can install the MCP and the skill with two commands).
Please check out this PR, there's a zip file with the skill that you can download..try the skill without the MCP and report if it works! 🧡
Please check out this PR, there's a zip file with the skill that you can download..try the skill without the MCP and report if it works! 🧡
feat: claude skill + marketplace by paoloricciuti · Pull Request #110 · sveltejs/mcp
This adds a svelte 5 skill that you can download manually to put in Claude code OR Claude desktop/web, but also makes this repo a Claude marketplace/plugin.
This will ease the experience of sharing...
github.com
November 14, 2025 at 11:40 AM
We are finally making a svelte skill for claude (and turning the repo into a claude marketplace so you can install the MCP and the skill with two commands).
Please check out this PR, there's a zip file with the skill that you can download..try the skill without the MCP and report if it works! 🧡
Please check out this PR, there's a zip file with the skill that you can download..try the skill without the MCP and report if it works! 🧡
Reposted by ratevyra
Welp, turns out it’s not real.
Just opening DevTools makes all timers 5-100× slower, due to the overhead of capturing stack traces. Even if you do nothing else (don’t record performance, etc)!
Guess who accidentally spent a weekend optimizing this 😅
(h/t @paul.irish for explaining why)
Just opening DevTools makes all timers 5-100× slower, due to the overhead of capturing stack traces. Even if you do nothing else (don’t record performance, etc)!
Guess who accidentally spent a weekend optimizing this 😅
(h/t @paul.irish for explaining why)
November 14, 2025 at 3:42 AM
Welp, turns out it’s not real.
Just opening DevTools makes all timers 5-100× slower, due to the overhead of capturing stack traces. Even if you do nothing else (don’t record performance, etc)!
Guess who accidentally spent a weekend optimizing this 😅
(h/t @paul.irish for explaining why)
Just opening DevTools makes all timers 5-100× slower, due to the overhead of capturing stack traces. Even if you do nothing else (don’t record performance, etc)!
Guess who accidentally spent a weekend optimizing this 😅
(h/t @paul.irish for explaining why)
Reposted by ratevyra
did a head to head, K2-T vs GPT-5.1, on a simple web search, and K2 is much better, more direct to the point, no fluff
November 14, 2025 at 1:16 AM
did a head to head, K2-T vs GPT-5.1, on a simple web search, and K2 is much better, more direct to the point, no fluff
Reposted by ratevyra
Have you used masking with gradients in CSS? What about repeating-linear-gradient?
I show a lot of neat CSS tricks while customizing a range input with no frameworks
youtu.be/AvvXZL3mKhc?...
I show a lot of neat CSS tricks while customizing a range input with no frameworks
youtu.be/AvvXZL3mKhc?...
How To Make Amazing Range Inputs With CSS
YouTube video by Syntax
youtu.be
November 13, 2025 at 3:28 PM
Have you used masking with gradients in CSS? What about repeating-linear-gradient?
I show a lot of neat CSS tricks while customizing a range input with no frameworks
youtu.be/AvvXZL3mKhc?...
I show a lot of neat CSS tricks while customizing a range input with no frameworks
youtu.be/AvvXZL3mKhc?...
Reposted by ratevyra
🤔 What if there were a conference without talks? Just hanging out with OSS maintainers & builders, seeing what they’re working on, and learning cool stuff by actually talking to people.
🤯 What if hallway track turned into the whole event!
💡 @tannerlinsley.com is exploring it. Interested? ⬇️
🤯 What if hallway track turned into the whole event!
💡 @tannerlinsley.com is exploring it. Interested? ⬇️
Web Forge Conf - A new kind of web developer conference
A community funded, not for profit event built for creators and the developers who use their work. Less stage. More story. All signal.
webforgeconf.com
November 13, 2025 at 9:36 AM
🤔 What if there were a conference without talks? Just hanging out with OSS maintainers & builders, seeing what they’re working on, and learning cool stuff by actually talking to people.
🤯 What if hallway track turned into the whole event!
💡 @tannerlinsley.com is exploring it. Interested? ⬇️
🤯 What if hallway track turned into the whole event!
💡 @tannerlinsley.com is exploring it. Interested? ⬇️
Reposted by ratevyra
The combination of the scrollbar taking space AND view transitions will be remembered as the most atrocious experience of the 2020s.
Motion sickness to the extreme.
Please use `scrollbar-gutter: stable`
Motion sickness to the extreme.
Please use `scrollbar-gutter: stable`
November 13, 2025 at 3:16 PM
The combination of the scrollbar taking space AND view transitions will be remembered as the most atrocious experience of the 2020s.
Motion sickness to the extreme.
Please use `scrollbar-gutter: stable`
Motion sickness to the extreme.
Please use `scrollbar-gutter: stable`
Reposted by ratevyra
We're filing litigation to completely dismantle "Lighthouse," a massive "Phishing-as-a-Service" operation that has harmed more than 1 million global victims. And we're also endorsing bipartisan bills in the U.S. Congress to protect everyone from scams. goo.gle/447Lltn
A dual strategy: legal action and new legislation to fight scammers
An overview of how Google is taking legal action, supporting strong bipartisan legislation and releasing new features to fight scammers.
goo.gle
November 12, 2025 at 10:41 PM
We're filing litigation to completely dismantle "Lighthouse," a massive "Phishing-as-a-Service" operation that has harmed more than 1 million global victims. And we're also endorsing bipartisan bills in the U.S. Congress to protect everyone from scams. goo.gle/447Lltn
Reposted by ratevyra
In a shift, the world will spend $40 billion more on new data centers this year than it will on finding new sources of oil.
Data centers now attract more investment than than finding new oil supplies | TechCrunch
In a shift, the world will spend $40 billion more on new data centers this year than it will on finding new sources of oil.
techcrunch.com
November 12, 2025 at 4:49 PM
In a shift, the world will spend $40 billion more on new data centers this year than it will on finding new sources of oil.
Reposted by ratevyra
MiniMax M2 — Open-weight and Agent & Code Native
⚡ Global FREE for a limited time via MiniMax Agent & API
- Advanced Coding Capability: Engineered for end-to-end developer workflows. Strong capability on a wide-range of applications (Claude Code, Cursor, Cline, Kilo Code, Droid, etc)
⚡ Global FREE for a limited time via MiniMax Agent & API
- Advanced Coding Capability: Engineered for end-to-end developer workflows. Strong capability on a wide-range of applications (Claude Code, Cursor, Cline, Kilo Code, Droid, etc)
October 27, 2025 at 11:31 AM
MiniMax M2 — Open-weight and Agent & Code Native
⚡ Global FREE for a limited time via MiniMax Agent & API
- Advanced Coding Capability: Engineered for end-to-end developer workflows. Strong capability on a wide-range of applications (Claude Code, Cursor, Cline, Kilo Code, Droid, etc)
⚡ Global FREE for a limited time via MiniMax Agent & API
- Advanced Coding Capability: Engineered for end-to-end developer workflows. Strong capability on a wide-range of applications (Claude Code, Cursor, Cline, Kilo Code, Droid, etc)
Reposted by ratevyra
Wikipedia, the online nonprofit encyclopedia, is urging AI companies to stop scraping its pages and to use its paid API instead.
Wikipedia urges AI companies to use its paid API, and stop scraping | TechCrunch
Wikipedia, the online nonprofit encyclopedia, is urging AI companies to stop scraping its pages and to use its paid API instead.
techcrunch.com
November 10, 2025 at 6:35 PM
Wikipedia, the online nonprofit encyclopedia, is urging AI companies to stop scraping its pages and to use its paid API instead.
Reposted by ratevyra
@pnpm.io getting better by the day
pnpm.io/blog/release...
I still can't believe that a one-person package manager is doing better than npm CLI, owned by a corporate, where the resources of the two projects are incomparable.
Draw your own conclusions.
pnpm.io/blog/release...
I still can't believe that a one-person package manager is doing better than npm CLI, owned by a corporate, where the resources of the two projects are incomparable.
Draw your own conclusions.
pnpm 10.21 | pnpm
Added support for Node.js runtime installation for dependencies and a setting for configuring trust policy.
pnpm.io
November 10, 2025 at 11:37 AM
@pnpm.io getting better by the day
pnpm.io/blog/release...
I still can't believe that a one-person package manager is doing better than npm CLI, owned by a corporate, where the resources of the two projects are incomparable.
Draw your own conclusions.
pnpm.io/blog/release...
I still can't believe that a one-person package manager is doing better than npm CLI, owned by a corporate, where the resources of the two projects are incomparable.
Draw your own conclusions.
Reposted by ratevyra
💯 this.
You are not only not helping, you are actively making it worse.
If you expect oss maintainers to go through the slop you generated without any work of your own, it shows how little you value our time and mental health.
This is even more insulting than the occasional rage post.
Do better.
You are not only not helping, you are actively making it worse.
If you expect oss maintainers to go through the slop you generated without any work of your own, it shows how little you value our time and mental health.
This is even more insulting than the occasional rage post.
Do better.
Vibe coding PRs - STOP please...
November 10, 2025 at 9:47 AM
💯 this.
You are not only not helping, you are actively making it worse.
If you expect oss maintainers to go through the slop you generated without any work of your own, it shows how little you value our time and mental health.
This is even more insulting than the occasional rage post.
Do better.
You are not only not helping, you are actively making it worse.
If you expect oss maintainers to go through the slop you generated without any work of your own, it shows how little you value our time and mental health.
This is even more insulting than the occasional rage post.
Do better.
Reposted by ratevyra
Vibe coding PRs - STOP please...
November 10, 2025 at 5:02 AM
Vibe coding PRs - STOP please...
Reposted by ratevyra
Every single time an LLM hallucinates, I am grateful:
Grateful that I spotted it, and thus remind myself that any and all LLM output needs to be validated. You can never trust these things 100%, unless you have additional validation in place that is 100% reliable.
Grateful that I spotted it, and thus remind myself that any and all LLM output needs to be validated. You can never trust these things 100%, unless you have additional validation in place that is 100% reliable.
November 9, 2025 at 2:20 PM
Every single time an LLM hallucinates, I am grateful:
Grateful that I spotted it, and thus remind myself that any and all LLM output needs to be validated. You can never trust these things 100%, unless you have additional validation in place that is 100% reliable.
Grateful that I spotted it, and thus remind myself that any and all LLM output needs to be validated. You can never trust these things 100%, unless you have additional validation in place that is 100% reliable.
Reposted by ratevyra
“I wish we wouldn’t tell people to follow their passion. I wish we’d tell them to follow their curiosity.”
— Astrid Bin
— Astrid Bin
November 7, 2025 at 4:25 PM
“I wish we wouldn’t tell people to follow their passion. I wish we’d tell them to follow their curiosity.”
— Astrid Bin
— Astrid Bin
Reposted by ratevyra
I just saw the Kimi K2 Thinking release!
Kimi K2 is based on the DeepSeek V3/R1 architecture, and here's a side-by-side comparison.
In short, Kimi K2 is a slightly scaled DeepSeek V3/R1. And the gains are in the data and training recipes. Hopefully, we will see some details on those soon, too.
Kimi K2 is based on the DeepSeek V3/R1 architecture, and here's a side-by-side comparison.
In short, Kimi K2 is a slightly scaled DeepSeek V3/R1. And the gains are in the data and training recipes. Hopefully, we will see some details on those soon, too.
November 6, 2025 at 7:35 PM
I just saw the Kimi K2 Thinking release!
Kimi K2 is based on the DeepSeek V3/R1 architecture, and here's a side-by-side comparison.
In short, Kimi K2 is a slightly scaled DeepSeek V3/R1. And the gains are in the data and training recipes. Hopefully, we will see some details on those soon, too.
Kimi K2 is based on the DeepSeek V3/R1 architecture, and here's a side-by-side comparison.
In short, Kimi K2 is a slightly scaled DeepSeek V3/R1. And the gains are in the data and training recipes. Hopefully, we will see some details on those soon, too.