Floor

@nahcrof @0xSero crof is pretty sick, I vouch

@0xSero crofAI has it, I can give you $1 credit to check it out if you'd like, I think other people have suggested it as well

Who has Deepseek-v4-pro on subscription right now? Preferably fast and with an api I can plug into all my harnesses

Kansas has 2 roadblocks to legalization; @TyMastersonKS and @DanHawkinsKS. #ksleg isn't even *allowed* to bring it up in committee. Just saying. #SaveKansas and never vote for them again.

so uh. hey guys. when are we doing a class action lawsuit on all the fake job listings. can we do that soon. or maybe pass some laws?

Crazy that we haven't even started the recession yet

@thsottiaux Its pretty cool, but even on low it took ~5% of my pro plan's 5 hour usage per request, so I did end up having to go back to 5.4. It seems qualitatively better though even on low reasoning @davis7's take about using the model on low seems ideal I think.

Have you tried GPT-5.5?

I'm planning on having deepseek-v4-pro up on crof by the end of the day tomorrow if possible (so within the next ~20 hours)

@LottoLabs the efficiency trade offs for their sizes would be awesome to compare if its somehow possible, the technical report mentioned how much more efficient V4 is compared to V3.2. Both the single token FLOPs and KV cache size efficiency gains DeepSeek found for V4 were impressive.

Deepseek V4 Pro High vs. the goat qwen 27b Surprisingly close among these benchmarks This either means these benchmarks are not representative of the models ability or 27b is just a dog

@banteg please tell me imagegen

I realized at 28 that I don’t want a ‘dream job’ I just want a life where I’m not stressed about money.

The fact Indeed put a daily apply limit, hides jobs based on your resume, and stopped showing how many applied tells you how bad the job market is fr

within the next 10-15 years, politicians are gonna run on confiscating the endowments of legacy universities who took six figures from kids and gave them worthless slips of paper

I turned down a 4th round job interview and 3 people called me to ask why. I told them their interview process is too long and it required me to take time off work each time.

Four points up and they’re panicking. This race isn’t about DC money. It’s about whether Kansas belongs to the people who live here… or the ones who fly in for photo ops.

You ever quit a job application halfway through because they’re asking too much?

@thdxr I'm not going to praise a company that seems like they mismanaged its compute resources and decided to push that problem off to their customers.

everyone clapping for dario being conservative with compute great example of how everyone is wrong about everything

@fleetingbits gdpr is probably the biggest one? I think their privacy and data laws are a lot stricter from what I've heard

why didn't mistral become a frontier lab?

@HealthRanger Open source is not that fair behind in most tasks. Proprietary models often have the advantage of a harness they are actively trained against, but they are hardly *definitively* better than Open Source models at this point. Minimax's M2.7 broke that threshold for me.

I'M CALLING THIS OUT: There is a concerted, top-down effort to dumb down all AI models that are accessible by the public, including open source AI and cloud-based AI. And there's a reason why this is happening. It explains why Opus 4.7 is such a disappointment, among other things. The chasm between in-house frontier AI / AGI versus open-to-the-public AI is widening. On purpose. I'll have a full report on this tomorrow or the next day.

2 weeks of content will come from this.

Can devs or researchers get sponsored by redbull? I really like the Cherry Sakura flavor.