lemonish97

4.5 is still the latest Haiku model

"It is nearly free on the max plans"

is a funny oxymoron

Looking at the next iOS rumors, I think it's inevitable.

They were mostly distilled or fine-tuned OAI models.

What is your evidence for this claim?

Isn't this exactly why Microsoft had to recall 'Recall'?

OpenCode and Github copilot are still options if you want the freedom to choose different models.

What's Cursor's moat here? I'm a bit surprised that xAI/SpaceX needs to buy them rather than building their own VScode forked IDE or an agentic UI/CLI.

I always see it as more of a social media company rather than a media co.

From the article: "Additionally, AI features in Notepad settings has been renamed to Advanced features and it allows users to toggle off AI capabilities within the app."

I honestly don't mind this, as long as it's not being forced. And I believe this feature exists only within their npu PCs.

I have a feeling it won't be this exact model, but rather smaller distilled variants, similar to the gemma line

The only solutions to anyone now are ad-blockers (that work) and premium. I still don't see the latter being successful, especially outside the US.

Try Playlet. Lookup the git repo. Has some really cool features like auto-skipping in-video sponsored ads too.

Also introduces offline mode for air-gapped environments

Agree. But you can keep the pelican theme in whatever new benchmark you choose to come up with. Iconic at this point.

Isn't their competition (at least in the mobile space) google/android themselves?

I hope they add a web search tool to the agent skills too. Most of my llm usage on my phone are just quick lookups and search summarizations. Would love to do these with a local model rather than Google AI mode of any other cloud based inference tools.

MAI-Image-2 is a really good image model. Seems to be doing really well on Arena AI too. Great to see msft reducing their reliance on OAI

unfortunately, that seems to be the eventual turn for any ai company, once they reach a user buy-in cliff

That's github copilot though, which is different (gh is msft, but here you have the option to choose other models too). Their regular copilot app and the one on windows still uses oai as their base model, which can hurt oai, if they choose to move away from it

Some of the nemotron models are really good. Hope this encourages more open-weight/source models from the west

They seem to be getting better or more powerful. The newer Intel Panther lakes and AMD Ryzen are over 50 TOPS now, IIRC

I wonder when we begin to see the dividends of all the NPU PCs come into play. AMD have been doing some good work with their NPU/iGPU hybrid inference kernels. If these larger models could be scaled down to run on NPUs, you'd see much better power advantages, compared to running them on the CPU.

[flagged]

Clearly shows you haven't opened the link?

"Working closely with Anthropic, we have integrated the technology behind Claude Cowork into Microsoft 365 Copilot."

Always love a new good SLM. Reasoning + vision under 20B sounds promising, will be testing this out

Honestly prefer teams over slack. Slack is good at text threads, while I'd still choose Teams for calls and meetings.

So an mcp server just for google specific public docs? Aren't there a dozen of them like Context7 (https://context7.com/) that already do this?

I know windows 11 is super buggy and riddled with issues (and the copilot mess), but I'm starting to feel there's a weird echo chamber around these forums that don't even bother looking at what the product or repository is, and automatically assume it's bad 'cause it's from Microsoft.

Or maybe it is just better?