A couple years ago I got back into watching competitive StarCraft again. Broodwar is still going strong in Korea, with some of the old names still near the top. If you're interested, check out the ASL English broadcasts on the SOOP YouTube channel. The finals of last season - Soma vs Flash - was an excellent watch. Artosis and Tasteless still do the casting, except they are now supported solely through Patreon and streaming rather than from the Korean broadcasting companies.
Tons of value left, especially for open source models. I would say the benchmark is yet to be truly saturated (just look at the legs and seat to see what I am talking about) and I always look forward to seeing them. Thank you!
i got a "dumb" sony one, the runt of its laser line but 5+ years later i havent changed it nor anticipate i'll need to, nothing better looking than hdr10 exists anyway, and the laser projectors are easily visible even in daytime with basic curtains down
requires either wiring on ceiling/wall and drilling or extra furniture under TV (short throw)
double power consumption over TV
the only benefit really is the size, but nowadays you can buy 65-75" TV which is already pretty huge depending on your room, so I fail to see argument for projector unless you need super large screen which you intend to watch from big distance/outdoors in very late evening
I encourage those who want to read more to start a book club with their friends. My club has read 6 books now and it gets at two things the article mentions, isolation and illiteracy.
I’ve recently also subscribed to a few in print magazines to address the desire for up to date recent information we get from social media.
HN is still hard to quit, the amount of useful information here is still high. digg.com summarizes X which I’ve found to be useful for not feeling like I am missing out on discussions there.
A book club of substance. We are awash in both high value books and pre and post AI formulaic slop. Any reading of books is probably better than no reading, but given some of the pop books, that might be debatable.
If you’re a copilot cli user (we exist) you know that the auto mode leaves something to be desired. It only changes models at the start of a session or after compaction. Most dev tasks are now sub agent heavy and there is no model routing on the sub agents.
I do notice that there is no mention of effort levels in the comparisons. 5.6 Luna Max is really good and really cheap. A 5.6 Sol high orchestrator with 5.6 Luna max is cheap, has frontier level performance, and is faster than Sol alone. This can be accomplished with simple agent instructions. Looking forward to running my own benchmarks on HydraFusion to see how it fares. Gone are the days of a single model doing all of the work it seems, unless the work requires no tool calls.
I'm not looking to change your view. But for other readers who are curious, here is a link to an interesting task to gain intuition. Ahmad is a good data point for someone who tinkered, built intuition, then started his own ai company.
https://twitter.com/TheAhmadOsman/status/2087742080793620593...
I think we are talking about different things to be honest. Understanding LLMs is a fine thing to do but to say you understand what the network is actually doing, even in toy examples like MNIST is extremely tricky. Figuring out what a ~3 trillion parameter network is doing in thousands of dimensions seems intractable to me.
Linking to a Twitter thread about training a small LLM to be good at Wordle is not an example of what I'm talking about. It might well be a useful task but it doesn't allow us to understand deeply what's happening.
What’s fun at the staff level is anticipating issues with the foresight that comes with experience. Going deep on a new tech so that when a situation arises, you are ready with the right tool and the depth to back it up. This is another way to deliver value the staff level.
They also make the orangiest orange paint, the yellowest yellow, the pinkest pink, etc. At a certain point I have to wonder how much these are accurate descriptors of the paint colors vs arbitrary made up marketing nonsense.
Note that there has been controversy for the last few years around Culture Hustle not fulfilling orders and doing other sketchy things. Just a heads up.
I had ordered something from them and it took me initiating a chargeback for them to ship anything at all. The product that arrived was of terrible quality, and
their customer support email is completely useless as well, any email gets a canned "your email is very important to us!" style response. My bank denied my second chargeback request because they had technically shipped something.
Having this work on Apple metal is huge for the accessibility around training splats. I remember two years ago I needed an RTX Nvidia GPU to run things like splatfacto on nerfstudio. Very neat optimizations as well like the depth map and sky dome, 3dgs has come a long way. Awesome!
Folding laundry seems like one but then quickly falls apart the more I think about it. Not sure what else.