will moss

1.7K posts

will moss banner
will moss

will moss

@thewillmoss

east nashville. software engineer. vandy alumn.

Nashville, TN Katılım Aralık 2011
694 Takip Edilen383 Takipçiler
will moss
will moss@thewillmoss·
@2024dion Scale is way off, but the solution is obvious. Gary-Chicago International Air, Space & Seaport
will moss tweet media
English
4
0
49
2.4K
will moss
will moss@thewillmoss·
@_fat_ugly_rat_ Growing up in the red area, I knew a lot of people who felt like Obama stole the nomination from Hillary. It feels foreign now, but the mid south still had a base loyal to the Clintons and Gores. Several of my teachers had photos of them meeting Hillary in their classroom.
will moss tweet media
English
4
3
64
23.8K
will moss
will moss@thewillmoss·
@rai__as @CollapseAnime For training AI models, you need the latest generation to keep up, but for running them you don’t. It’s why X is using older chips to run Anthropic’s models, but newer chips to help Cursor train composer
English
1
0
2
95
will moss
will moss@thewillmoss·
@rai__as @CollapseAnime Samsung and TSMC only have so much capacity, so they focus on the latest generation. Chinese firms can’t produce latest gen chips, but they are getting really good at making last gen chips
English
1
0
2
98
Crispr (Third-Worldist) 🌍📉
China's GPU being $400 USD and the performance of a 3070 isn't the point (you can replicate this with a 5060). The point is they're catching up and with their own domestic ram production, they'll likely make models that are way less constrained on ram as well.
English
26
119
2.1K
28.2K
will moss
will moss@thewillmoss·
@thdxr Maybe in a few months when Effect is in more training datasets and reasoning gets better they might be able to use Effect better in greenfield cases without examples
English
0
0
0
91
will moss
will moss@thewillmoss·
@thdxr Yeah, I think the best packages and languages for agents are still the ones they’re trained on. Javascript and Python have benefits, but they’re also the majority of the training data. They can make up for reasoning ability by basically auto completing
English
2
0
4
3.8K
dax
dax@thdxr·
effect is actually bad for agents
dax tweet media
English
33
1
697
73.1K
will moss
will moss@thewillmoss·
@Prototype27F @The_Devin_Green Imagine being a Betazoid when the Pheobian navy attacks. Only the nomadic betazoids still left on the great beta steppe are safe
English
0
0
30
2K
will moss
will moss@thewillmoss·
@JupiterKunt @_fat_ugly_rat_ This is the kind of medieval construction project that would have caused someone to get the nickname “Harold the Builder” or “Wilhelm the Great”
English
0
0
13
324
James🗳
James🗳@_fat_ugly_rat_·
Category: Cities in the Americas with geography that would have gone insanely hard in medieval Europe
James🗳 tweet mediaJames🗳 tweet mediaJames🗳 tweet mediaJames🗳 tweet media
English
74
161
4.7K
637.2K
will moss
will moss@thewillmoss·
Anyone else use a non-trivial amount of tokens on doing git operations they don't want to do by hand?
English
0
0
0
52
will moss
will moss@thewillmoss·
@loveof2is1 @gfodor Yeah, if you ask a dozen agents the second one, all of them can reliably give the correct answer. Some “thinking” agents can even make the process worse, because they over think what the first one is even asking for
English
0
0
2
120
will moss
will moss@thewillmoss·
@thdxr I tried DeepAgents, Claude, and OpenCode’s SDK, but OpenCode was the only one that could work with Kimi+GLM+MiniMax plans reliably
English
0
0
1
71
dax
dax@thdxr·
i'm working on some sdk design for v2 what i keep coming back to is not fully understanding one segment of the usecase every infra company is prepping for "millions of agents to run on us" are you working on a product that needs this? tell me more
English
40
2
270
27.9K
will moss
will moss@thewillmoss·
@techfrenAJ @calirenato82 @opencode @geminicli @qwen @grok I really like the website. I've been looking at a lot of different repos trying to establish a standard, but all of them seem abandoned and/or no stars on gh. I wish we could get some bigger voices like opencode to adopt it
English
0
0
1
50
cali (renato caliari)
cali (renato caliari)@calirenato82·
I built agent-sync because I was tired of asking myself: - Are my skills in sync across @opencode , @geminicli , pi and @qwen qwen-code? - Did I backup my configs? - What if I lose all my custom skills? After several commits, here's what I learned 🧵
English
1
1
2
230
will moss
will moss@thewillmoss·
@meganysta @levantophile I’ve done a lot of deep wikipedia dives on Louisiana history, and I’ve visited a few times, so I’m glad it came in handy lol
English
0
0
1
16
T. ☀️
T. ☀️@levantophile·
Put aside the thick American accent, this man’s French sounds closer to the French spoken in France today than the dialect spoken in Quebec does. I wonder why Quebec, which was lost to the British only a few decades before Louisiana was sold to the United States, seems to have preserved a more archaic form of French, whereas in Louisiana, the local dialect seemingly continued to evolve alongside the French spoken in the metropole. Also, it’s impressive that French-speaking Louisianans managed to preserve the language for more than 200 years as a small minority within an overwhelmingly English-speaking society.
Ceri⚜︎@cherrysinceri

Un vieux monsieur à Lafayette, Louisiane (avec Jourdan Thibodeaux)

English
29
62
1.1K
118.9K
@jason
@jason@Jason·
We started an AI founder twitter group... reply with "I'm in" if you're a founder and want to be added
English
10.8K
134
4.6K
904.6K
Sarah Fong
Sarah Fong@MilksandMatcha·
Giving away 5 Windsurf Max ($200/month) plans Each person will get 3 months of free Windsurf Max (highest tier). Try out SWE 1.6, Cognition's latest, fastest, and most intelligent model, powered by @cerebras. Winners will be selected from comments in 48 hours, comment below why you want it.
Cognition@cognition

We’re releasing SWE-1.6, our best model in both intelligence & model UX. SWE-1.6 matches our Preview model on SWE-Bench Pro while dramatically improving on various behavioral axes. It’s available today in Windsurf in two modes: free tier (200 tok/s) and fast tier (950 tok/s).

English
1K
51
855
162.4K