Can you please add a model selector? I already have gemma-4-26B-A4B-it-QAT-MLX-4bit on my MacBook, and I see Gemma 4 26B is one of the models that will be downloaded automatically for my 64GB machine.
Maybe that's true, and maybe you're an outlier. But from what we've been seeing, the frontier LLMs have torn apart many codebases, finding bugs littered everywhere, some over a decade old.
At this point it seems impossible for human to write better/more efficient code than a swarm of agents -- to be clear, multiple coordinated LLMs, not one-shot builds with a single agent.
Ten median US software developers cost about $1.5 - $1.7 million per year - including all benefits - as of a couple of years ago.
A frontier AI is at least as capable as ten median human software developers. It's likely well beyond that now. Over the course of a year, would you have to spend $1.5 million to get the output of those ten median humans? No chance in hell. You'd have to spend less than 10% of that.
Then on top of that: the AI is improving rapidly; the humans go on vacation, the humans decline, the humans get old, the humans get sick, the humans take parental leave, the human struggle to adapt, the humans are resistant to change, the humans riot, the humans protest, the humans need healthcare, the humans want paid more, the humans can only work for 4-6 hours at peak while writing software (slowly).
Do you know any humans that are experts at a dozen programming languages? That human does not exist. The frontier AI are omnimaths.
What's my cost as a business to utilize the latest cutting edge Chinese model, renting its usage from the cloud? Whatever the Chinese cost (on their side), I'm never going to suffer even a tiny fraction of that pass-along cost no matter what. We can all appreciate the subsidization going on.
Or to grab a bunch of $200 per month accounts from OpenAI or Anthropic for a team to use?
The calculations is: I'm in the US, if I have ten median software developers, what's my cost? Then versus using AI instead.
The human labor can't keep up. Even if we freeze things right here, and then hyper optimize Fable 5.2 by etching it straight to silicon, the human developers are obsolete. And the future isn't going to stop coming.
I like their angle though - from a societal perspective of allocated resources, how long until we've paid off the cost of building these LLMs? Imagine instead the entire supply chain was Stalinistically tuned towards finding bugs in Linux. Training rare earth mineral miners how to code level or whatever. I think as a thought experiment it's interesting, the answers could translate to an organizational level about comparing labor cost against token cost. What's a good breakdown of token spend vs hiring some new juniors to train up?
Maybe at Opus 4.6 level for the "easier" tasks. But honestly if you were given a choice between a computer with Qwen 3.8 or one with even older Sonnet 5, both local, same tok/s, window size, etc... what are you picking?
I've got QFN making a C(sub/superset)->microcode compiler, writing it in OCaml. It's come up with optimizations that Claude didn't come up with (in the 4.6/4.8 era) when it was making the previous C-based implementation. The compiler generates microcode for an algorithmic state machine/transducer in addition to Verilog. I've been working with QFN on this for about 4 days now and it's pretty close to feature complete - yesterday it was able to compile our Game of Life C implementation and the microcode ran on a Tang Nano 9K FPGA board with LCD displaying the cells. I do not think this is an "easier" task.
Similar story here, except we went from Jira to a custom Plane.so build. Everyone from product to engineering loves it, gets features they want added quickly. And for me, I mainly have Claude keep track of everything (mcp).
The Jira billing administration will remain as a business school case study of dark patterns for years. Can't wait for the EU to put their nose into it.
I've had the idea for years for a browser extension that showed positive/negative sentiment scores next to each post, based on comments. Mostly because often the top comment on posts are "ackchyually..."
Intel has supported native hardware decoder for AV1 since Tiger Lake. That’s quite old. You could buy a used NUC with Tiger Lake. Even when new a NUC with 1115G4 or 1135G7 isn’t expensive. I found one with damaged WiFi for $20. https://www.ebay.com/itm/366767313411 You can buy it now.
I went to the link to buy a computer (a NUC11TNKi3) with this capability for $20, but all it would let me do is place a bid on it; I cannot buy it now.
Cheaper to get a long fiber optic HDMI cable and connect your TV to your PC GPU. That's what I do. You get the best uncompressed video (typ 4K 120Hz or 8K 60Hz), zero lag so it can be used for stuff other than video, and don't have to maintain another box
I tried this and it was a huge pain. This works in a small place with a single TV and a single user, past that it's rough.
Running a hdmi cable through walls sucks, you're limited in tenancy (I can't game while wife watches the TV), controlling the pc remotely is annoying, most streaming apps don't support 4k on a pc (same problem with a nuc), I can go on.
This does not pass, or get anywhere close to passing the wife test imo.
Lots of media devices that don't have AV1 decoders tend to have issues decoding high resolution AV1 in real time so you either get pauses to buffer or dropped frames.
Trip planning is the most often used use-case for agents/bots I see in marketing. I'm lucky to travel once a year, but if I did have a trip, I wouldn't feel comfortable letting an LLM take the wheel. Too many variables for wrong flight or room, closed locations in itinerary, etc.
Now, critical server infrastructure? Sure, have at it Claude, make no mistakes!
I would literally never let an AI plan my trip. I do have very specific interests though and it's nice to have an overview of suggestions of places to go.
Plus, it's great at researching the history of places you're visiting, so you can find cultural experiences a bit easier.
I don't travel often, but I do live in Europe where it's easy to visit a new city or country in the weekend. This makes this functionality a quite common use for me.
https://thefirmwarehub.com/
reply