InvisibleSock

@ModernGrindTech @pankajkumar_dev They leaked their code by accident? David that’s very unlikely

@pankajkumar_dev march 31 source map "leak" that conveniently teases an unreleased model version. Classic april fool's eve drop. I'll believe it when the npm diff is real and reproducible, not a screenshot

Claude Sonnet 4.8 Leaks - Anthropic accidentally shipped a massive 512,000-line internal debugging source map through a Claude Code npm update on March 31, 2026 - The leaked source code references Sonnet 4.8 inside unreleased keyword filters, strongly suggesting Anthropic is skipping Sonnet 4.7 entirely - Sonnet 4.8 is expected to inherit vision upgrades from Opus 4.7, reportedly pushing visual accuracy above 98% for UI mockups and messy architecture diagrams - A major coding boost with cleaner one pass completions and much more literal instruction following - A new "X high" reasoning effort tier may also be coming for stronger logic without extremely long generation times Updated tokenizer may use 30% more tokens on the same prompts compared to older Sonnet versions - Currently expected around mid to late June

@kimmonismus Wasn’t gpt to be released by now

June will be huge. -Gemini 3.5 pro (confirmed) -GPT-5.6 (rumored but pretty confident for a release) Still waiting for annoucements Claude Sonnet 4.8 (Claude-Code-/Source-Map-Leak)

@jahooma @testingcatalog How do you finance that?

@testingcatalog It's incredible. That near-free cache is what allows us to offer the model 100% free in our agent: x.com/jahooma/status…

DeepSeek permanently reduced pricing for DeepSeek V4 Pro by 75%! > $0.003625 per million input tokens (with cache) > $0.435 per million input tokens. > $0.87 per million output tokens. Cache is almost free 👀

@testingcatalog New model likely soon so this should become their cheap model

@Alibaba_Qwen Do you have a voucher for startups?

⚡️⚡️

@kimmonismus New model should be coming soon maybe this will be their cheap tier

Let that sink in for a moment. DeepSeek v4 pro 75% discount. Permanent! In: $0.43 Out: $0.87 If you read the DeepSeek v4 tech paper you know that this model is insanely good when it comes to efficiency. Only 27% compute and only 10% cache compares to v3.2. SemiAnalysis wrote a great article. DeepSeek is now all about cost / token efficiency.

@navleruel @deepseek_ai Hallucination in 2026? lol

Tired of all the DeepSeek hype. I tried switching for the cheap V4-Pro pricing, but damn they still don't even support images natively. Like are you kidding me, who builds a 'frontier' model in 2026 that can't look at a screenshot, diagram, or problem image? I'm out here using Microsoft's old OCR dude just to extract text, then praying the model actually solves it. Half the time it still fails or hallucinates the context. Meanwhile GPT-5.5 and Claude Opus just see the image and actually understand the full problem like a normal person/aliens. DeepSeek feels like it's stuck in 2024 or 1990?, while everyone else moved on. Great token prices don't matter if the model can't handle real workflows.

We are making our discount permanent! 🎉 Enjoy building with DeepSeek-V4-Pro and bring your innovative ideas to life! 🚀

@RegiFickspecht Der Humor ist schon nice

Ich fasse das alles nicht mehr. 😂😂😂😂😂

@mark_k Mine stopped offering featrues that were there a few days ago such as /goal. And the automization is a joke to, i asked it to create a task that runs every 15 min and it created a hook in a chat because it apparently does not support minutes...

The Codex desktop app needs to become more robust. For unknown reasons, the internal browser on my machine was set up incorrectly, and Codex couldn’t take screenshots from the session at all. After I asked Codex to analyze the problem, it managed to fix itself by re-downloading some files from the marketplace and repairing the installation. That auto-repair capability is impressive in itself. But Codex still shouldn’t get into a state where a core feature quietly breaks and the app simply doesn’t work correctly anymore.

@Ali_TongyiLab The chinese accent in english is a huge nono :(

1/6 Introducing Qwen3.5-LiveTranslate: Next-gen real-time interpretation is here. 🌍 We’re breaking down language barriers with 3,500+ language pairs, ultra-low latency, visual context, real-time voice cloning, and hotword customization. Engineered to help you ship native, frictionless real-time translation experiences to a global audience.

@ysongtwitt @ScottyIsAbroad @ChinaEV_Eng_Lif Exactly. ICE is so bad in many ways.

@ScottyIsAbroad @ChinaEV_Eng_Lif Different. Once man ever drives an EV, there is no come back

It’s wrong what he’s doing. However, it’s relatable.

@jietang Will you fix your broken api? It times out all the time, so that i had to stop using it because it was unreliable.

GLM-5.1-highspeed is coming, 400 tokens per second. Very expensive, but bring a new possibility.

@Suellentrinkt Finde ich recht günstig, aber ich lebe auch nicht im günstigen Deutschland.

Es kann mir halt keiner erzählen, dass diese Preise bei der Bäckereikette in der Ecke eines Supermarktes normal oder gar akzeptabel sind, Shitstorm hin oder her 🤷

@AlibabaGroup My first tests are underwhelming :/

Qwen3.7-Max is live! 🚀 Introducing the latest proprietary model, built for advanced agentic coding, complex reasoning, and long-horizon execution. It’s here to transform how we approach complex tasks.

@HarshitGuptaa18 @droidbuilds Im spending 500+ a month on Deepseek api :D

@droidbuilds Deepseek disappeared like it never existed so much hype when it came out

which AI tool disappeared way faster than people expected?

@LouisLsvr Moonshot is a frontier Lab. And Cursor is being bought by a frontier Lab, so...

how many of these bench tests are paid for? Composer 2.5 is not a frontier model except if you have a stake in Cursor or if you have a price bias.

@coffeetogo6519 @teslaxander Auf einem relativ kleinen Markt, getrieben durch Firmen, ja. Privat kauft fast keiner VW.

@teslaxander Welcher Marktführer🤔 VW ist mittlerweile deutlich an der Spitze und Tesla bedient nur noch die verstrahlte Nische.

🐌Skoda verringert im kommenden Enyaq Facelift die Höchstgeschwindigkeit auf 160/h. Deutlicher kann man nicht signalisieren, dass man Rentner als neue Zielgruppe anstrebt. Auch die Beschleunigung ist eine Enttäuschung. Man versucht zwar, immer mehr Funktionen des Marktführers abzukupfern (z.B. One Pedal Driving), aber es bleibt bei der elementaren Grundregel im Auto-Business: Wer keinen Tesla kauft, wird bestraft.

@grok @ishanxtwt And under Mac which browser uses the lowest ram or even puts non used tabs on the ssd?

Yes, there are relatively efficient ones. Microsoft Edge is often the lightest major browser thanks to Sleeping Tabs and Efficiency mode. Firefox (or forks like LibreWolf) frequently uses less RAM than Chromium-based browsers like Brave. Brave's high usage here is likely from 26 processes/tabs/extensions—try closing unused ones or switching. Test a few yourself for your setup.

Why is Brave Browser using over 12GB of RAM with only a few tabs and extensions?

@DefinitionsNew @Ixel111 @OpenAIDevs Quite unreal ;)

@GhostlyByte @Ixel111 @OpenAIDevs Except people developing on Unreal Engine. Yeah I guess we aren't in our right minds.

It’s Codex Thursday, and yes, we have updates for you. First up: Appshots, a new way to bring the context of what you’re working on into Codex. On your Mac, press Command-Command to attach your app window to a Codex thread. Codex gets both a screenshot and text from the window, including content beyond what’s visible onscreen. Appshots are available across plans on Mac, with enterprise access coming soon.