tipsytoad

I’m not sure I understand the hype around this model. Isn’t this just an llm with a chat template, with the options prefix cached?

  <option>option A</option> <option>option B</option><endofoptions>userprompt<eos>

Then the llm is constrained to a few special tokens indicating the possibilities? e.g. <option1> <option2>

Plus it takes 3x the wafer capacity for hbm than the same byte capacity in dram, so we’ll likely see the consumer market be decimated here

seems unusably slow, and is this for short context?

idk I thought the same until I got personal amex support at work, now it’s my roofline on how useful/trusted an agent could be. Now I book all my travel through them, and it’s an absolute blessing

Is LinkedIn a normal social network? It seems like an absolute dumpster fire to me (it even has shorts now!)

I’m quite happy that my friends have labelled me as “an absolute idiot with anything but computers.” It’s much easier to relate to people at that level, adopt a happy-go-lucky attitude, and being wrong about 99% of things. being the smart one isn’t such a great label, and is (mostly) all in your head

It’s 27b active params vs 13b active params, so you’d expect it to be 2x more expensive when serving multiple users

And the goalposts must move once again..

I think this comes in with the wrong assumptions from the start. The thing that US vs Apple taught us is to not demand access publicly, this puts companies in an awkward spot. If approached more tacitly, gag order etc the company has nothing to gain but everything to lose.. and more likely to comply. I hugely doubt that intentional backdoors don’t exist for the most powerful countries / people

Bit of a counter point, but I ran a very nice wfh setup at a previous job, 4k monitor and ergo keyboards. I was a bit dismayed when I joined a new co and found that we had 1080p monitors and a generic Logitech keyboard. After a week of the text looking pixelated I entirely got used to it and think maybe complaints about monitors being a health hazard and rsi are a tad overblown by a vocal minority.

Stylistically ugly code that is globally consistent (ideally across the entire ecosystem) >> locally beautiful code that is inconsistent with the rest of the codebase / up to authors taste. No one likes black formatting, but it’s at least consistent. “Your car can be any color you like as long as it’s black”

With regards to your example, adding a comma to the final element should preserve multi-line. You just don’t know the formatting rules yet apparently

Yeah came here to comment exactly this. And this is generally why I dislike/avoid this type of first principle analysis: it can make very convincing arguments that are just totally wrong due to some misleading assumption

1.0 is actually pretty arbitrary and way too high as a general rule. Something like 0.3 is a more sensible default

And yet the monologue is a complete work of fiction, a script delivered by a talented actor that we still find moving. So what are these authentic experiences to you, or does it not matter if we can’t tell the difference?

Curious how they do a “blind” preference test. To any evaluator I’m sure it’s quite clear which answer is AI vs human.

Who designs a website like this?

We can use jj now, thank goodness. But I still miss my old git workflows + lazygit

I joined gdm recently, and previously used (neo)vim exclusively. Begrudgingly Cider-V is very, very good. It might be possible to get by without it, but the system is so locked down you’re going to make a lot of sacrifices. (very few authorised extensions, codebase is so large it’s going to break whatever tools your used to using anyway, no git)

I’m well thinking I may as well trade my brick of an m5 pro for a 13” chromebook, it’s a strange time.

As someone who basically only hangs out on HN, what are the trends on Reddit? I thought it would vary by channel

Throughput is a metric for the total number of tokens/sec for all users in the system. Latency (ITL,TTFT) are individual user metrics.

The guys previous post was how rabbit r1 is revolutionising the smartphone. So I would take this post with a grain (heap?) of salt

Made it to the 5th stage of grief

While I large agree, when I rely too much on agentic llm usage I come away feeling that I haven’t really learnt much over the session, and the code wasn’t really “mine”. It’s also easy to let your skills atrophy over time if you’re not careful, and for the hardest / interesting problems I often turn the llm off entirely and write out the code by hand, and come out a lot happier than just guiding Claude

Someone not familiar with the field rediscovering the stochastic parrot argument from 3+ years ago

Clearly not from the UK. By US standards Labour would be socialist, and conservative (right) liberal at best.

It’s a quite deceptive paper. The main headline benchmarks (math500, aime24 /25) final answer is just a number from 0-1000, so what is the takeaway supposed to be for pass@k of 512/1024?

On the unstructured outputs, where you can’t just ratchet up the pass@k until it’s almost random, it switches the base model out for instruct, and in the worse case on livecodebench it uses a qwen-r1-distill as a _base_ model(!?) that’s an instruct model further fine tuned on R1’s reasoning traces. I assume that was because no matter how high the pass@k, a base model won’t output correct python.

I get the same feeling that I'm "not being productive" while playing video games, watching tv, etc that seems to kill any enjoyment from doing these things.

For me learning piano has been a great alternative to programming in the off hours (typing is quite transferrable too!). Highly recommend if you're like me on screens all day.

Like, PyTorch? And the new Mac minis have 512gb of unified memory

I usually am a huge fan of “copilot” tools (I use cursor, etc) and Claude has always been my go to.

But Sonnet 3.7 actually seems dangerous to me, it seems it’s been RL’d _way_ too hard into producing code that won’t crash — to the point where it will go completely against the instructions to sneak in workarounds (e.g. returning random data when a function fails!). Claude Code just makes this even worse by giving very little oversight when it makes these “errors”

I wholly disagree with the comic, but a anti AI art take I’m more sympathetic to: https://x.com/soi/status/1815584824033177606?s=46