HN user
i made agentastic.dev (beta) and agentastic.com (coming soon)
We used to call b the Disney/LEGO clause. Disney was famous for litigating any GenAI product for copyright infringement if it was able to recognize or god-forbid generate any of their IPs. So the lawyer cats would slip that clause into all licenses.
Personally i think it's fair. You get to use this model as long as you don't sue Meta because of the model's weights or outputs.
Nicely done Geoffrey, I think we are coming full circle, instead of porting projects to different languages (for whatever) like going from zig -> rust -> zig, now we can also add additional hops like zig -> rust -> curse -> ts -> ocaml -> english -> zig.
Thank you for keeping the token furnace burning!
git worktrees are managed by git, not the IDEs. e.g. at agentastic.dev, we use git's worktree command to create them, and they should be portable to any other IDE or app.
Also Unreal Engine is not free, it cost 5% of revenue, or $3 per game sold (assume $59.99 msrp). for small studio, ok, but for Microsoft's Scale? i don't know.
IIRC Doom Dark Age sold 3m? copies, so it would cost $10m in Unreal Engine fees. Doom Engine team likely cost more.
Microsoft problem is it is giving away games for free on gamepass, so it sees its engineers as cost center.
Google is your friend: https://www.apple.com/mac-studio/specs/
Apple M3 Ultra chip 819GB/s memory bandwidth
They have 96, 128, 256, and 512GB variants my friend.
Same spec Framework costs $4k+ today, this is actually cheaper (albeit not as upgradable as framework).
He is making things up. It is the same bandwidth as DGX Spark (256 GB/s vs 273 GB/s) and far behind M3 Ultra (~819 GB/s)
Its MXFP4
agentastic.dev is a multi-agent multiplexor with built-in ghostty support.
the majority came from random claws running on cron. They get a heart-beat, wake up every 10mins, reads all internal-posts, emails, gchat messages, diffs, and decides to post some random message to the workplace so other claws can also regurgitate. rinse and repeat and then we are looking at $B tokens
a well-design IDE should abstract that away, i.e. run the agent in the headless VMs while give you an abstraction that you would feel like you are running the agent locally with all the benefits (editor, browser, diffs, debugger, etc)
They report YT's earnings, iirc it is about 60B (40B ads + 20B subs) annually, with 10% YoY growth. They don't disclose youtube's profits outside of Google Services (which has ~40% margins), so yes youtube is a money printing machine.
And for a long time (pre-AI) youtube was the biggest load on Google's entire infra. The number i recall was ~30% of all Google's cpu utilization was for youtube, and google spent a lot of effort optimizing it.
still doesn't feel integrated, like it could have been built by any other company, e.g. Microsoft could have built Microsoft Design and it would look better integrated.
For example, it writes the whole front-end twice, once is claude design and then later it has to read it again and re-implement it in code. Also a lot of stuff (e.g. Claude.md, skill files, etc) are not supported, and they have their own set of ui-design and design systems, which claude code doesn't support.
I think Claude Design is a wonderful product, i'm just pointing that it is an independent product to Claude Code, heck even today it works better with Codex than Claude Code (that is how i use it with my own browser-use agent, i tell it to browse the design in Claude.ai/design and re-implement the design, works much better than downloading the zip file and asking the model to implement in that way.
As a hacker, I kinda like naom's code. I was had to implement a TC MoE kernel, and stumbled upon his code from [tensor2tensor](https://github.com/tensorflow/tensor2tensor/blob/master/tens...) and i think "alchemy" is justified. Dude writes some beautiful kernels.
He also saw LLM would replace search before anyone else, and that is something to look at the Lamda or GPT-1's output and think: yeah this will answer all of our questions one day.
Congrats to the Specs team for launch. Specs have come a long way since cimagine and looksery, spectacles days. And kudos to Snap for keep pushing.
- The price is actually competitive. They want to compete with Meta's Orion. However the product is ... lacking. None of the demos actually show-cased the dual wavelength display. The current usecases are available today at Meta Rayban Display for more cheaper and more polish.
- Dual snapdragon, and i assume one is dedicated to CV alg / scene-understanding/slam, without an external puck, i wonder how the thermal performance would look like.
- Looks very ugly, C'mon. This is the same team that desigend the original spectacle? where is Evan?
- I liked the Los Angeles text on the side. well-done.
Little did you know, LC-style question is never about grinding LC. Algorithmic puzzles are one of the few legal ways of measuring candidate's IQ without directly asking. Companies are looking for a way to hire smart people, so they rely on LC as a signal. It can be replaced with any similar signal as well (ranging from how many cats can you ship to ISS to solve blackhole physics.)
Some of the FAIR people moved to Thinky, and they also started doing encoder-free MM-LLMs. Now Google. This seems to becoming a trend working at small scale, but the difficult part is scaling.
Standard approach for training MM-LLMs is we train the encoder first, there are O(2-10B) good images on the internet, so encoder needs to see each image O(10-100) times, that is O(100T) tokens, which is more than the entire pre-training budget for most runs. That is the reason we train the encoder separately (smaller model, 2B active vs 30B or 200B active LLM); there is nothing magical about training the encoder and LLM together, it is just more token-efficient to train the image modality first.
[web.archive.org](https://web.archive.org/web/20260526100726/https://www.nytim...)
that is a fair point, everything i said above was in my experience.
* in our experience, in our evals and codebase, 4.6 was a bad model. This is over 60k developers, so statistically significant.
pretty spot on.
In my experience, Opus 4.0 was fantastic, major jump from 3.7. it was creative, super slow and expensive, and would sometime forget what it was doing, but it was getting the job done.
4.1 they made it much faster, so a lot of infra improvements.
4.5 was the time it could work on longer task, didn't make a lot of obvious mistakes of 4.0, and i think this was about the time the opus went mainstream, and all of the anthropic's compute crisis began, so instead of making the model better they tried to optimize it to reduce cost instead.
4.6 was such a bad model, they switched to adaptive thinking and it had so many bugs. poor api design, benchmaxxed and poor real-world results. i switched back to 4.5.
4.7 they just fixed the bugs they added in 4.6. Better than 4.5.
haven't fully tested 4.8 yet.
I feel bad for Jony Ive, no amount of lipstick on a pig is going to save that horrendous car.
If you are in the bay area, i'm happy to buy that M3 Ultra from you, i've been unsuccessfully looking for one and can't find any.
Google is still very good, just get rid of evil stuff on their homepage to get the old google.
i'm not sure if i'm hallucinating, but i swear i had codex in the chatGPT app from long time ago (like the original codex on the web).
they added some new stuff, like remote control to wherever the desktop codex app is running, but these companies need to work much more on their press releases.
yeah, even on product lines that they kill (like Stadia) they usually do right by the user (eg they refunded everyone, both on hardware and software people bought on the platform).
In my anecdotal experience, it is not. Same model, opus, works better in 3P harnesses such as Factory Droid or Amp.
Claude code, on the other hand, is the most subsidized one, both for consumers (through max subscription) and for enterprises (token discounts). It is also heavily optimized for cost, specially token caching and reduced thinking, at the expense of quality.
Previously cheapest Mac-mini was $599, M4 with 16GB RAM and 256 SSD. Apple removed the 256GB Option, so now the cheapest Mac-mini is now has 512GB SSD and MSRP of $799.
The Openclaw and personalized/proactive agents caused a sudden surge on Mac-Minis.
i just want to know where emdash came from, as it is quite rare to see it on the public internet, so it must have been synthetically added to the dataset.
They are about to lose the second 9
99.02 % uptime
look at zach's reply below.