uf00lme

its due out tomorrow right?

Cursor is saying we need to get improve the harness and tooling, but the models are getting better. One would have to assume that will continue and that cheaper/smaller models will continue to become even more useful.

This tends to be the kind of thing you can get thrown in for free when negotiating. It's highly unlikely to be a deal breaker for anyone buying Perforce, if anything they may have done this to stop annoying cheap customers buying from them.

It's surprising usable and cheap enough to run in 'fast' mode when vibing something quick. For simple code I find I prefer the code it writes over GLM or Gemini family.

You can thank the us gov for a list of companies to invest in.

I wonder why they didn’t train off Kimi 2.6, I hope is it because they already had a good base and not that they messed up that relationship.

Woah, is this part of the future of models? Basically little models you can use as tools.

The speed is impressive, I wish it could be setup for similar to speculative decoding

A 2011 MBP is likely a better a general purpose PC, those early models had some great engineering. Wait for the reviews and benchmarks but the M1/M2 based MBPs are still great daily drivers.

Channels matter a lot, quad channel ddr4 is going to beat ddr5 in dual channel most of the time.

Do you know which MRI you used? Not all are equal. Most MRI are 1.5T powered, and you can’t get fine details until you hit 3T. And there are differences even in the 3T power range. There are higher powered MRI which are mostly only used in research, whilst it is a bit scary thinking about the sheer power of them but a 7T machine doing a full scan of you, would be guaranteed nearly to find anything wrong with you.

When I last looked the full body scans for sale seemed to used 1.5T setup, which seems like a waste. The 3T advanced scans looks much more detailed, but it just depends on where you live - I couldn’t find any around.

I think having kids asking if they can scan the face of their parents/older sibling will probably give rise to some good conversations.

Mirrors are not effective enough. Shielding drones from energy weapons seems like a similar problem to entering Earth’s atmosphere, you want to shield it in a way that will blast away safely and ideally diffuse the laser, so the energy is spread over a larger space. I suspect larger lasers will likely aways win, since there is only so much shielding can do. At which point we could end up with transformers like drones that are built to be broken apart mid flight and yet still deliver damage. I feel like defending drones could become possible with energy weapons but only under ideal weather conditions.

It might feels natural to try and use terraform to deploy kubernetes resources since you’ve likely configured the cluster using it, but the helm/kubeneters/kubectl providers are limited by terraform’s way of working. So whilst the providers try to marry the two when deploying anything complex it generally ends up feels like a hack and you lose a lot of the benefits of using terraform in the first place.

In my experience, it’s best to bootstrap ArgoCD/flux, rbac and cloud permissions those services need in Terraform and then move on to do everything else can via Kustomize via gitop. This keeps everything sane and relatively easy to debug on the fly, using the right tool for the job.