seb0

🚦 Traffic CV & geospatial analytics @ platomo
OpenTrafficCam • MOT • trajectories • AI agents

[seb0 @seb_ox]

I'm pretty sure this is meant to say Fable 5 and not Opus 5.

[seb0 @seb_ox]

@theo Yes

[seb0 @seb_ox]

@mitchellh @thsottiaux Agreed, it’s so annoying to switch to the chat mode when it constantly flips to work again when I just want to use Chat.

[seb0 @seb_ox]

@kimmonismus EU suddenly becoming a good place to use AI again lol

[seb0 @seb_ox]

@GonGonsior @claudeai Yes it did.

[OpenAI Developers @OpenAIDevs]

We’re bringing Codex and ChatGPT together in one desktop app.

The same powerful coding agent, now alongside ChatGPT Work, with new coding workflows, a new Chrome extension, revamped in-app browser, and faster Computer Use powered by GPT-5.6.
https://t.co/9hOtxyDEw5

[seb0 @seb_ox]

@kaaaash____ Probably manufacturing limits. The 15 Pro and 16 Pro only had titanium on the frame. The 17 Pro is a full aluminium unibody which is probably harder to pull of with titanium.

[seb0 @seb_ox]

@suni_code They wrapped other labs AI models in their search engine and had to pay full price for their API. Now, others are taking the crown, such as @ExaAILabs.

[seb0 @seb_ox]

@claudeai Cursor has just posted their benchmarks.
To me, it looks like the real reason for the Sonnet 5 is to compete with the GLM 5.2 at a lower price.
https://t.co/Oon8kLcTVx

[Cursor @cursor_ai]

Claude Sonnet 5 is now available in Cursor.

On CursorBench, it's a meaningful step up from Sonnet 4.6: 57% vs. 49%.

[seb0 @seb_ox]

@claudeai So it looks like in many cases it is actually cheaper to run Opus 4.8 on low than using Sonnet 5 or am I getting the chart wrong? 🤔 Not sure if the chart’s cost axis is using the cheaper introductory pricing or the standard ones.

[seb0 @seb_ox]

@rules_business @claudeai Haha same, Opus is forever stuck on xhigh for me too. My guess is that Sonnet could be cheaper for subagent work or for the dynamic workflows stuff. Not sure if Agents can even set the effort level of subagents though because then low Opus could be better again 🤔.

[seb0 @seb_ox]

@zudasworld @claudeai Could be, but the chart implies this is the cost per task. If Opus 4.8 can finish the task with fewer tokens on low effort than Sonnet 5 on xhigh, I don't see any real value. I might be wrong or missing something.

[ClaudeDevs @ClaudeDevs]

Claude Desktop is now available on Linux (Ubuntu and Debian) in beta.

Alongside the browser and terminal, you now get a first-class desktop experience with Claude Code, Claude Cowork, and chat on all paid plans.

[seb0 @seb_ox]

This is maybe a bit dramatized but I still agree with this take. Europe 2031 — What getting AI wrong means for us https://t.co/piQBbYGisQ

[seb0 @seb_ox]

@gdb Would be nice if mortal people would get access to this...

[seb0 @seb_ox]

@Polymarket I don't belive this is achievable. It's against the law of physics.

[seb0 @seb_ox]

@TechByTaraa I don't think a switch is necessary. It makes more sense to combine both and use them in conjunction to achieve maximum quality.

[seb0 @seb_ox]

@theo @FrameworkPuter Your Hermes agent decided it was time for an upgrade 😉.
Btw, I love the Framework Desktop. It’s a real amazing dev server. I use it every single day!

[seb0 @seb_ox]

@cheatyyyy That crazy expensive, but they also host it crazy fast...
I just set up Qwe3.6-35B on my RTX 4090 and it’s blazing fast too, with up to 170 tok/s with 256k context.