What's interesting for me is that it shows related papers where arxiv's influence flower/CORE have none. Only bummer is the few papers I've tried have no comments lol (the two are maybe correlated?)
Nearly a year later and Khruangbin's A LA SALA still is a top 3 album play and one of the best of the 2020s so far
Love this framing of AI-assisted dev. I've been thinking for a week or so now that our small (3) dev team needs a shared prompt library to generate consistently. We're all using Sonnet3.5, but in slightly diff ways. cursor.directory et. al. seem like good starting points.
They're at another off-site reimagining the future 😝
I'd use it! Totally agree, though it seems to have levelled off the last month or so
Joe's idea to keep a tree of markdown feature docs/progress is also neat. Gonna give that one a try
Yup, looking forward to reading! Thanks @qianli.dev!
An example I see: sharing b/w Cursor and an otel platform. "We learned that this operation isn't thread safe, and we haven't found all the places it's used yet", then in the otel tool "I see some strange thread behavior, perhaps its related to this lesson we learned?"
I think of MCP as equivalent to LSP/DAP protocols. If we can expose that cursorrules file to more applications that LLMs have access to then those lessons can be shared across apps.
Agree with your other thread re:memory! The protocol is just a spec for how LLMs interact with a standardized set of http endpoints. We're getting into xkcd/standards territory here but I think there's value in having services that we can leverage across apps.
I've thought that this would be a great use case for model context protocols with a memory server. Then you'd have a portable set of requirements across models and clients rather than just the code editor. Not sure if cursor supports/will support mcp tho
Curious to hear how it goes for you. We're heavy on lambda/sqs right now but I never got into step functions. DBOS seems like it's a simpler take from some brief browsing
The prompt engineered approach is likely useful for a variety of requirements, not just perf opt. I've had good success w/ the LLM by having it output code that answers a series of questions as well. Excellent read!
As a first time viewer I liked the format! Yes, micro-blogging is nice when you have the time to compose it, but it's nice to also have the variety of video content. Thank you!
Like bringing up a set of applications, or a quick app launcher? Curious to see what else you get up to with it... I've always struggled to find a place for those kinds of tools in my workflow. I think the closest I've come to that sphere is Alfred workflows heh
In fact, the only time I really strayed away from vim et. al. was when we were both in the Activity space. LSP/DAP impls weren't quite mature enough at the time to supplant jetbrains, but I suspect things like metals is far enough along now that you could get away with it for JVM dev
IOW I wouldn't necessarily switch now if wezterm is working well for you haha ghostty is early days (despite the 1.0), we'll see if long term it has legs.
Hammerspoon I haven't given a go yet, but I'm obviously biased about neovim 😝
fwiw I am still satisfied with kitty! The only thing that was going to get me to switch was rendering/navigating large blobs of text lol
Now that I'm here, I realize that I can have a pretty nice replacement for my tmux workflow as well, so I've finally simplified down to fish+neovim+ghostty
Lot of hype around ghostty. Other than sane defaults, biggest change for me is seeing big perf increase while scrolling large tsx files in neovim. Kitty struggled, ghostty has no problem.
Jury is still out on whether it warrants the hype 😝
Sounds like a good time :D I keep hearing about serverless on non-aws platforms but it's hard to leave old faithful Lambda haha.
Ceci n'est pas un livre :^)
Watcha reading?
Anyone in the #neovim sphere working on a model context protocol plugin? I want avante.nvim to leverage mcp servers. I tried wrapping the python sdk in a rplugin but the experience of installing/updating leaves a lot to be desired. Writing a lua mcp client seems like the next best option?
CSR and SSR; project is <1 year old
Caveat tho the container is 3gb+ in size, so if you're going to be frequently updating it then make sure to prune the older images. I found out the hard way that colima has an 8gb disk size default this way lol
If you want to dip your toes in the water before integrating with ollama directly open webui will give you a chat experience similar to chatgpt in the browser. They've been adding a ton of stuff like tool use and agents that I don't personally use, but it's still a nice OOB experience.
Ollama has done well for me. I currently run an M3 w/ 32gb of unified memory so 7b models are doable with decent t/s. 30b+ models require quantization that compromises quality in a lot of cases.
This is still on my plate but had to divert attention to some work stuff haha
Since it's applied to all follows it's a bit of a sledgehammer if you just want to do it for a few folks haha but it's the first thing I thought of
Are you familiar with the Only Posts feed? I think this does what you're looking for:
bsky.app/profile/did:...
Hah yeah, we took turns cause my hands kept cramping
A helping hands solder station holding wireless headphone charging kit. Wires are delicately touching the positive and negative terminals usually used for the headphones
A helping hands solder station holding wireless headphone bud. Wires are delicately touching the positive and negative terminals usually used for the headphones
Discharged these headphones too much to the point they are undervolted. Taking the extremely reasonable step of macguyvering a manual charge until they have 3v to start the normal charging process 😒😒😒 the last pair of helping hands are my partner's hah. What an adventure.