mackross

I'm setting up https://github.com/tencentcloud/CubeSandbox on my homelab to give it a run against this. It apparently supports e2b-dev out of the box which is quite nice.

Can’t wait to use this. Amazing work.

I wouldn’t at all be surprised if OpenAI settled with HF to not press charges.

Supertonic 3 is pretty amazing too. I’ve got time to first sound down to well under a second by streaming chunks and stitching together from CPU with goonx (non c-go onnx) - happy to upload model if any go devs interested.

What a happy surprise today! The amount of times I’ve had to do weird janky package APIs so the API was still reasonable is more than I can count.

moved to kobo (the nice one with colour screen) with calibre web running behind a cloudflare tunnel, getting books direct through smallest publishers/authors. Adds the tiniest bit of friction in book acquisition but reading experience, battery life, everything else shits on my kindle experience (and I’ve owned every one).

The antigravity teamwork-preview doesn't work for me -- upgraded to ultra, installed antigravity 2, ran teamwork-preview, keeps failing: "You have exhausted your capacity on this model. Your quota will reset after 0s."

Durable objects looks interesting! Thanks for the link

I’m a huge elixir fan, but imho it doesn’t solve durable execution out of the box which is a major problem that often gets swept under the rug by BEAM fanboys. Because ETS and supervision trees don’t play well with deployment via restart, you’ve got to write some level of execution state to relational database or files. You can choose persistent ETS, mnesia, etc, (which have their own tradeoffs and come with some kind of gnarley data-loss scenarios in deep documentation). But, whatever you choose, in my experience you will need to spend a fair amount of time considering how your processes are going to survive restarts. Alternatively, Oban is nice, but it’s a heavy layer that makes control flow more complex to follow. And, yes you can roll your own hot code deploy and run in persistent VMs/bare metal and be a true BEAM native, but it’s not easy out of the box and comes with its own set of foot guns. If I’m missing something, I would love for someone to explain to me how to do things better, as I find this to be a big pain point whenever I pick up elixir. I want to use the beautiful primitives, but I feel I’m always fighting durable execution in the event of a server restart. I wish a temporal.io client or something with similar guarantees was baked into the lang/frameworks.

- A phoenix/ecto inspired batteries included framework for Golang. Uses data-star for real time bindings (can do live view like things but my personal favorite is just real time form validation out of the box). Hot reload with templ, daisy, and tailwind (no npm required). Routes file provides metadata on routes so type safe route helpers are generated for views and handlers.

Same name as my Phoenix inspired framework for go: https://codeberg.org/lixgo/lix

Cool app. I couldn’t see a way to report an error in one of the default expressions.

Guess my edit didn’t work…

An often overlooked extra advantage to Google is their massive existing ad inventory. If LLMs do end up being ad supported and both products are roughly the same, Google wins. The large supply of ads direct from a diverse set of advertisers means they can fill more ad slots with higher quality ads, for a higher price, and at a lower cost. They’re also already staffed with an enormous amount of talent for ad optimization. Just this advantage would translate into higher sustained margins (even assuming similar costs), but given TPU it might be even greater. This plus the gobs of cash they already spin off, and their massive war chest means they can spend an ungodly amount on user acquisition. It’s their search playbook all over again.

An often overlooked extra advantage to Google is their massive existing ad inventory. If LLMs do end up being ad supported and both products are roughly the same, Google wins. The large supply of ads direct from a diverse set of advertisers means they can fill more ad slots with higher quality ads, for a higher price, and at a lower cost. They’re also already staffed with an enormous amount of talent for ad optimization. Just thus advantage would translate into higher sustained margins (even assuming similar costs), but given TPU it might be even greater. This plus the gobs of cash they already spin off, and their massive war chest means they can spend an ungodly amount on user acquisition. It’s their search playbook all over again.

Love the thought put into mise and now fnox. They’re a joy to use.

Amazing work by author!

Always enjoy mattt’s work. Looks like a great library.

Very cool :) can it just do observability too or do you have to use for all prompting?

We’ve built a solution in conjunction with a university to this problem that is pretty low effort to implement, but very few professors can be bothered to even try it out (the apathy and red tape is unreal). Honestly, it has been disheartening that distribution is so tough, as the results have been great for those who are using it.

I love this — it captures what I’ve been struggling to articulate after using o1 a lot.

I used these in a distributed sync in about 2013-2014. When Postgres got JSON support we ended up going centralized for that product and a simple logic clock was all we needed. Nevertheless, I still think they’re very cool.

Still possible, propshaft works perfectly with the official js-bundling and css-bundling gems which let you add any js build pipeline as a build step

We built a simple but novel solution for that is far more reliable and works completely differently to gpt-zero and openai methods. I'm not posting a link as we're not ready for HN hug of death, but please PM if interested.

The saddest thing is that this project has been once of the most demoralizing I've ever taken on. Day to day we see so many students being failed by teachers and school leadership who care more about "adapting to AI" than real student outcomes today.

In practice, we've found teachers don't generally want to have the difficult conversations with students when the hard evidence of cheating is given to them.

And generally school/university/college leadership have no real tactics to implement their "AI strategy" other than train their own chat bots (wtf) and "adapting assessment to use AI".

Unfortunately, a simple non AI fix to the problem is definitely not as good for their careers.

IMHO without a change, we're creating a pretty bleak future for students of the next few years.

I find programming outside-in ends with a better design and is generally faster than inside-out. Similar experience with the rest of the advice.

Exact same thing happened to me

Nice library

This should be taught early on to programmers. This post also reminds me of Ian Coopers talk on how TDD is misused at NDC.

https://youtu.be/IN9lftH0cJc?si=7h2IHnJyUNSU7Vrl

Another BedJet fan. My wife and I both agree it’s in the top 3-4 purchases since we’ve been together (10 years).

I think I would give up most other possessions before it. It is probably only beaten out by a few very sentimental items (like my wedding rings), Kindle Voyager (by far the best eReader Amazon ever made IMHO) and a couple of my newer Apple devices (Apple Pencil 2 and AirPods with iPhone are just magic).

Individualized climate control in bed is a luxury that has put a grateful smile on me for almost every day over the last 6 years.

Do you use the BedJet sheet? I find it much nicer than the air blowing directly.