jeffyaw

m5 max with UD-Q4_K_XL and 260k context is what i'm working with. and using compaction push it past 260k context just fine. i found when limiting the context that was one of the bigger debilitators. although, your setup sounds obviously better.

disagree, it’s absolutely comparable. but an m5 max is where it shines. and there are 50 flags to finetune with llama server before you get it right. i do prefer llama server over mlx currently on mac as well.

if you want to try a new runtime oam.js is focused on TypeScript & MCP servers.

i've been using a snapdragon x elite and loving it, albeit on Windows. 32GB ram

you can use Kimi K3 on the typed++ model tier: https://typed.cloud

this sounds like you're stuck in 2010.

it's not that they can't afford to buy one. it's that they're not available. and now a lot of other tlds have gained traction so it's now reasonable to consider others.

thanks for clarifying.

so who/what countered Linus on this and where is the discussion or final conclusion?

once you have built up a trust in a model/harness/setup/workflow and you’ve been hitting enter for months straight then Auto Mode is appealing.

i found initially it would hang on some commands (moreso on subagents and moreso when first rolled out) and then roll past the hung command sometimes forgetting about it entirely. which wasn’t cool.

but after dialing in subagent permissions and probably some updates from anthropic Auto Mode is great. i still don’t always do it. but after reading this article just switched my current 12 panes to Auto.

appreciate the reply!

just made oam.js and released it yesterday. check er out.

https://oamjs.org

optimized for typescript and running js/ts mcp servers.

could you help me understand the appeal of running everything in tmux? i use tmux if i have long running scraping/db backup or restore, or something similar.

it seems like an awesome thing to geek out on.

but i keep claude code sessions running for days that sometimes get interrupted that are easily resumed. without the tmux latency or giving up aspects of rendering.

why even take funding at all?

completely agree! not interested in getting funding at all.

thought about it and kept remembering that scene from Silicon Valley, “Nobody told me I could take less!”

i was an early vagrant user, a long time tf user, and 3 year nomad/consul user. but have moved on. cf now for aws & gcloud cli in scripts for gcp. and eks/gke instead of nomad.

An AI CLI for coding - more volume per dollar, cheaper extra usage, and monthly billing instead of 5-hour reset windows. One curl installs the launcher; then run t (or typed) from any project directory.

i will be digging into the CLI option, kudos.

you're right, i remember pre 4.6 i believe 200k was the norm.

highly agree, i built typed for this exactly.

playwright is AWESOME for user testing (works great with electron!).

it is not, it just helps to not go in the wrong direction as much. even in my comment i mentioned reviewing the code multiple times. so i do agree reviewing is essential.

minimax m3 has a 1M token context window so not sure how op is hitting this 200k. maybe the plan they're on? or some setting in some layer of whatever their dev tooling is using.

if you want to try minimax m3 with a lot of optimizations and a custom tui check out typed. and for that harness you can use it in yaw mode.

you can also use within claude code tui by running: typed cli off

i should probably change that to typed tui off/on. anyway.

try using a harness. i provide one in Yaw Mode and you can copy it and modify and use a modified if you wanted to learn and tweak yours.

i use the skills /yaw-review excessively sometimes multiple times in a row on the same pr or session. followed by most often /yaw-address-all and then /yaw-coverage to add tests and /yaw-ship-ready to make production ready.

after a few rounds of these they are not needed every time on the same codebase.

if you are desperately wishing programming to go back to the before times it will never. or it will always be there but expect to be incredibly less productive than your peers.

absolutely, for me the tui, ultracode agentic workflows, and streaming logic are far superior. the closest model is minimax 3.0 imo and i ended up adding a custom tui, agentic workflows, streaming logic and implementing skills to that (in typed) in order to get to an acceptable claude fallback. on their own i haven’t found one model comparable to claude, not even chatgpt.

i bet 10+ things you used today were authored by claude in one place or another.

february was some kind of nirvana. i do think claude code versions and what is introduced at that level is/was relevant.

but 4.8 xhigh w/ ultracode to me is just about Fable level (w/ some agents harness tweaking).

but have to switch to 4.7 xhigh and 4.6 max quite often these days.

i just use review skills on my sessions/prs extensively until they're in a good place. then coverage skills (which add test coverage).

i built in skills to work with M3 in a service called typed, an ai cli. it uses m3 under the hood (up to ~500k tokens), then switches to deepseek for up to 1M. a few bells and whistles added of typescript/python coding optimization. and just built a custom TUI frontend for it (initially works with the claude code tui and still does).

to toggle the typed tui you can run:

typed cli on

typed cli off

there is, yaw terminal.