skolos

This is discussion and not mechanical - surely I would understand this and suggest second favorite from not long ago.

I did lots of interview on both sides and know how often interviews are just luck - got asked trick question that I knew answer to, coding problem that involves algorithm I just recently used, etc.

There is so much randomness that I, as interviewer, decided on the following strategy - put candidate into best possible position and judge from that. I settle on one set of questions - "tell me about your most favorite project, why, and let's discuss in detail". If candidate knows his/her stuff - this is fun/informative discussion. Also easy filter if a candidate cannot say much or doesn't understand details of project they consider their favorite.

1 line in CLAUDE.md I tried (one line, 100 lines, many iterations) - it not always follows style instructions.

The value is very large for me - instead of reading page of Claude produced prose, I mostly just look at focused choices and continue with single button press - significantly less Claude deciphering and typing now. In addition, I do my sessions with /rc so I get nice form on mobile app where I can just select a button instead of typing to continue work.

PS: I think there is confusion in term - "every turn" - I don't mean every tool call, I mean after minutes or hours of agent work when agent decides that its done it gives a report. The report I see now has structure to it with options set to continue (or stop).

What to trust? Closed hardware/software that goes to trash if something goes wrong, or opensource that you can send an agent to debug if there are issues? I'm not trying to diminish JetKVM's product - for $35 its a steal if you don't want to bother with DIY and time involved, but I'd strongly disagree that opensource is less trustworthy than closed product.

or make one yourself: https://github.com/espkvm/espkvm

If you have same schedule every week, don't go on vacation, your kids don't have breaks, don't have seasons, then 'set and forget' dumb thermostat will work for you without any problems. For the rest of us we need some 'smartness' in our thermostats.

it is, unless it set to xhigh - it really likes generating tons of tokens for its thinking. unfortunately, for decently reliable coding results you want it on xhigh ...

There was a study specifically related to Qwen3.8 27B that showed that kv cache quantization has almost no impact on this model all the way to q4:

https://arxiv.org/html/2609.04098

On many models that I tested in past context quantization had very bad effect on model performance. However qwen3.8 27b is different.

I'm now running NVFP4 quantized both weight and cache on my RTX5090 and getting excellent results: 264k cache allocated for pool, 10k tok/s prompt processing, 200 tok/s generation for single stream, or 801 tok/s generation for 8 concurrent streams. Also have about 2Gb vram left for use of OS.

my coding agents regularly reach 200k context used without noticeable degradation.

P.S. I used setup from: https://github.com/seanyourhighness/vllm-sm12x-nvfp4-dflash2

why so many people add 'please' when asking machine to do something? Was there actually research that when you SCREAM or curse it follows your instructions better?

P.S. Although my wife insists that I should stay polite in case AI overlords remember how I treat them ...

Interesting - in my setup (llama.cpp rtx5090 qwen-3.6 27b) prompt processing with mtp is almost half vs non mtp. Sounds like I need to investigate what is wrong.

"Multi-token prediction gives a free speedup of up to 2x on many models" - at the expense of halving prompt processing speed

deterrent - no one starts war with a country that has nukes

There was lots of discussion within Russian and Ukrainian war analytics that nukes (at least tactical) are useless in this war for the following reasons:

- they would not change much on battlefield - there is no large concentrations that you can nuke - everything is dispersed

- nuking urban centers again won't change much on battlefield but would alienate China

- Russia's equipment is known to be not most reliable/maintained and worst that can happen to Russia is them trying to nuke and nukes not working

It sounds more like market based allocation than gameification.

I'd say adding another 16Gb gpu would be worth it - you'd be able to run larger model/larger context all within gpu's. It would give you more options of what you can run fast. Your current model probably doesn't run completely from GPU (depending on quants I don't think you can squeeze Gemma4:26b into 16Gb vram), so you already have some layers running on gpu and some on cpu. If you add another gpu you might be able to move all layers to vram which should speed up things for you. The layers calculations happen on whatever gpu's it sits, so the layers that are already on your rtx5080 would compute same, but the layers that currently your cpu handles will be computed with faster vram/compute of rtx5060.

> The Department of Defense said the xAI data center powered by the gas plant is critical to national security, revealing Grok was used to fire thousands of missiles in the Iran war.

How many times did you try? Same model running multiple times can produce both very good and very bad results. In my benchmark even 10 runs often not enough to tell for sure if one model is better than another.

I like Mathematica and use it regularly. But I did not see any benefits of using it over python as a tool that Claude Code can use. Every script it produced in wolfram was slower with worse answers than python. Wolfram people are really trying but so far the results are not very good.

Claude code regularly asks me questions - I like how anthropic implemented this

Looking at examples I see:

``` c = Sector(radius, start = 180°, end = 270°).translate(y = radius); ```

Programming language that requires (maybe it does not require, but then example is not good) to type degrees. Or maybe it is not designed to be typed and rather ai generated?

They did operate on a "continuous refresh" basis. However, it mostly stopped for almost 2 years now. Other than HW4 I don't think anything else is different between current models and their iterations 2 years ago.

Edit: mostly speaking about Model Y, as Model 3 had actual refresh recently.

Here you go: https://blog.adaptiverisk.com/post/2023-08-22-dataslicer/ Calculations are local, not in the cloud.

> shit build quality

That's old news. If build quality is your only concern, I suggest you to check them again. They supposedly improved build quality significantly after initial rollout of M3. As a data point: my family owned and drove 6 different Teslas over last 3 years - we did not have build quality (or any other) issues with any of them. All recent horror stories you heard are because of current Tesla scale and media negative bias against the company - you don't hear similar stories from other manufacturers.

Last two years used Tesla prices were higher than new ones. You need to wait up to a year to get new Tesla though. So you could actually make some money while driving newest versions of cars.

> the company optimizes delivery numbers of cars sold by sacrificing quality control

How do you know this? Did you see their internal data? There are lots of anecdotes going around about Tesla's quality. However, with all the TeslaQ it is hard to believe that there is real correlation between anecdotes and data. Here are my anecdotes - I owned 6 Teslas over last several years. Not one of them had any QC issues. I had one service done because I hit tire debris and front break dust shield started making noises. Tesla fixed that for me quickly with no charge. As for the data - during earnings calls they mentioned that they do pay close attention to their customer experience data and they had period of time where service was lagging. But they started addressing this issue and saw improvements. The way they are growing I do believe they need to keep close eye on customer experience, but looks like they understand that themselves and use data to make sure they are on top of this. Unfortunately there's not much reliable independent data to have better understanding of this issue.

Add: The intent of my comment was to ask if parent info is based on specific data or just anecdotes. As an example, I gave my own anecdotes and mentioned that they are not reliable correlation to the data. Somehow the responses I've got are all about anecdotes, mine or others, also some personal judgement of my ability to appreciate cars or judgment of my life circumstances that required me to have these many Teslas. Can we get back to discussing the main point I'm making - do we have data to make any of these judgements?

It does work: NIO does it (they have 700 battery swap stations already). CATL also announced that they will produce swappable car batteries.

I, myself, don't see how this can be more competitive than superchargers. But I do see that some customers would like to have this option.

They did try it in 2013, but abandoned the idea: https://www.tesla.com/videos/battery-swap-event

How Russian people would know about this? You are underestimating efficiency of Russian propaganda machine. Russian government does not shy away from inflicting damage onto Russian people and pointing finger to the West.

Interesting point. How would you resolve this chalantly? 144 million people who probably will struggle to purchase their next iPhone vs 44 million people many of whom might die, loose all property and homes. What is a good way to resolve this in your opinion?