Chromadera (@Chromadera) header
Chromadera (@Chromadera) profile picture

At the precipice of success Local ai Qwen-Lover AMD Rocker

Joined May 2023
Photo by Chromadera (@Chromadera)Photo by Chromadera (@Chromadera)Photo by Chromadera (@Chromadera)Photo by Chromadera (@Chromadera)Photo by Chromadera (@Chromadera)Photo by Chromadera (@Chromadera)Photo by Chromadera (@Chromadera)Photo by Chromadera (@Chromadera)Photo by Chromadera (@Chromadera)Photo by Chromadera (@Chromadera)
Chromadera retweeted
there are so many inference providers, so so so many they've all acquired some amount of compute, nothing crazy and most of them are not very good at inference and have no ability to directly generate demand so it's kind of a waste
45
6
2
417
17,044
15s one shot with pika h3 flex. currently using the loras , wondering what the diffusers would look like fully.
2
3
114
asked for slightly more consistency and got this. havent done the voice loras and setups, this is just plain pika flex h3 six steps at 960x544. about 10 minutes for each
19
Chromadera retweeted
LFG!
Replying to @nothingto12
Hey! We are looking into this. Sorry about that.
1
9
375
Chromadera retweeted
my $30 @bot agent made me $2.1k on prediction markets this month (proof below) and i only spend around 30 mins a day on it i built the model once, now i just tell the bot to update whenever i need it it pulls past market data + the latest news on X, then gives me the stats and projections i just double check the thesis, size my positions and make the trades research that used to take me a couple hours now takes minutes still a lot to improve, but it's already paid for itself there's plenty of room to scale will share the workflow soon 👀 no larp, proof below
7
1
2
13
433
Really Good. But would be nice if we can continue a conversation we started on the pc , from our phone
官方版适配移动端在即,我却打造了一个DeepSeek Harness移动端的终极答案?! 地址:github.com/ccch1mneyyy/dsh-g… 精美原生UI+手机本地运行DSH+远程控制电脑+插件市场+社交功能+宠物系统+可交互UI+手机操控+可更换主题+兼容官方所有插件++++… 随时随刻,想Go就Go~
1
1
116
Chromadera retweeted
Round 4: Ready for something different? Let's run SWE Bench on each provider and see how close they are together! Will DeepSeek API, Opencode Go, OpenDesign and Ollama have the same numbers? Let's find out.
🛎️ Round 3: DeepSeek API vs OpenCode Go vs OpenDesign vs Ollama. Now in the Pi harness! Who do you think did better? Rune Compiler is a puzzle game where players connect runes to build small programs. It tests whether a model can: - Build a working compiler, simulation, and 3D interface - Follow detailed rules across eight puzzles - Use automated tests to find and fix errors Ollama 25:39 wall time, 3.84s mean TTFT, 7.58s p95, 14.39M tokens, 0 inference errors DeepSeek API 44:06 wall time, 4.24s mean TTFT, 7.58s p95, 24.75M tokens, 0 inference errors OpenCode Go 70:19 wall time, 8.08s mean TTFT, 7.57s p95, 28.14M tokens, 0 inference errors Two first-token waits of 195s and 256s raised the mean OpenDesign 119:37 wall time, 5.76s mean TTFT, 9.88s p95, 31.63M tokens, 3 inference errors Two connection errors and one billing-scope timeout Demos are available here: nazzysar.github.io/deepseek-…
1
1
31
996
Chromadera retweeted
Why is my feed so suddenly full of political stuff? Where are my graphs and charts? I guess I'll have to make up for it with a hype reel
10
5
3
204
4,352
Testing out AMD and Strata today with 64GB DDR5 and a 9060XT
18
7
1
172
6,699
The thing I find jarring the most is people are finding it hard to believe that an employee works within your residential compound does not make you their employer or automatically entitle you to their labour.
So the rest of you with cars can't increase your own payments? I don't even get the logic behind this complaint.
1
2
304
Chromadera retweeted
Waiting for Qwen4😎. Heads up: I’m sharing my local Qwen3.8-Flash Q6 + Strata service for 5 hours. If you’re unsure about the Q4/Q2/Q3 quants you’re currently running, give it a try yourself. Give this to your agent setup for ya:
2
2
1
10
2,048
Okay the pika h3 flex is pretty sleek.
2
201
Chromadera retweeted
There really are people who post the most slop nonsense on a literal timer to fish engagement. Especially in AI/tech. What's your content? What is YOUR opinion? What's your niche? It's just "cold take x, what do you think?" Kinda lame that I gotta see this in my timeline.
2
1
8
556
lol this one still lives with his/her parents
2
2
167
Bruh has an item worth 30,000$. Spending $100 yearly to offer a degree of protection for that item is farfetched to you.
Lmao people really dey this world sha 😭
1
121
Chromadera retweeted
Well, that was sudden! Thank you all for 5K! Especially thank you Theo for the shout out! I think I'm gonna do some time per task charts tomorrow, since there are many different benchmarks that show time per task and they don't all agree gn y'all!
This dude is the only person I’ve seen actively comparing actual dollar values and speeds across AI subscriptions in a realistic, accurate way. Instant follow.
33
4
2
283
12,755
Chromadera retweeted
User funds are safe. Binance will cover all hacking losses. Deposits and withdrawals are temporarily suspended while our team conducts security checks, freezes funds, and works with law enforcement. We will share a full post-mortem once the investigation is complete. Please rely only on official Binance channels for updates.
1,085
977
401
7,776
913,013
Chromadera retweeted
Really tired of local LLM benchmarks with 8k/32/64k etc. Bench it at min. 128k - 256k.
115
56
29
1,258
50,801
Chromadera retweeted
The future of Africa can’t be cross border payments mehn
1
1
6
234
Chromadera retweeted
2 months after Fal said they're release H3 Max, it remains unreleased Meanwhile, Pika took the opportunity to release a H3 fine-tune that performs better than it:
Replying to @KeyTryer
we will release the weights, model keeps improving
8
14
2
292
15,342