Minty

🕵️‍♂️🔎 ai / ai Agents / Vibe Coding 🤖👨‍💻
Crypto & Prediction Market Enthusiast

[Minty @MintyNFT]

@TheMattBerman @openclaw Ugc

[Minty @MintyNFT]

@MitcheIl AGENT

[Minty @MintyNFT]

@shoaibyazdani @openclaw Ya it was an issue with my llm provider kimi coding. They flagged my account. Banned my account for use with openclaw. Switched providers and all is good.

[Minty @MintyNFT]

Just ran the @openclaw update 2.17.26 and now it doesn’t recognize my api keys. What do I do?

All models failed. HTTP 401: User not found (auth)

🤯 will have to wait until I get home to figure it out.

[Minty @MintyNFT]

@LottoLabs Running 35b, my 27b 4bit runs at 5-8tsp. 😭 need that fast 27b. Cant afford it.

[Minty @MintyNFT]

@kimmonismus 35b runs 7-8x 57tsp faster than 9b 7-8tsp for me. 🤣

[Minty @MintyNFT]

@Zeneca 🤣🤣🤣🤣 always breaks down.

[Minty @MintyNFT]

Buy or don’t buy a couple of used rtx 3090s?

I’m running qwen3.5 27b opus and only getting 8-9tsp. Not fast enough for local openclaw use.

can’t justify the price.

Will handle local coding, simple tasks and openclaw heartbeats.

[Minty @MintyNFT]

@kirillk_web3 lol this is bs. That video is of someone else not a Chinese university programmer. The videos is from a known quant ai agent coder @MoonDevOnYT 🤣

[Minty @MintyNFT]

@Cnicholasdownie @LuxAlgo I thought luxalgo only provides the data, mostly Indian brokers. It’s because the plugin runs in your ai interface like openclaw, opencode, claudecode. You can just do the same without it.

[Minty @MintyNFT]

@shydev69 @Immortalshiv1 I run 35b 4bit on m3 36gb at about 50-60 tps.

8B fp16 (full model) runs at 8 tps 🤣

[Minty @MintyNFT]

anyone compare Qwen 3.5 35B A3B 4 bit (20.4gb) vs Qwen 3.5 9B bf16 (18.8gb)?

55 (35B) vs 8 (9B) tokens per second

comes down to speed vs minor diff in size. hopefully even at 4bit 35B comes out significantly ahead.

mlx on mbp m3 36gb.

[Minty @MintyNFT]

@VadimStrizheus same running it.
maybe beats any claude opus/sonnet/haiku below 3.5?

[Minty @MintyNFT]

Holy shit, Qwen 3.5 9B is better than Gemini 2.5 Flash Lite!

I've been trying to get my local llm setup right to lower my openclaw api wastage on crons, heartbeats, etc.

can run the full model (non-quantized) on my laptop.

[Qwen @Alibaba_Qwen]

🚀 Introducing the Qwen 3.5 Small Model Series
Qwen3.5-0.8B · Qwen3.5-2B · Qwen3.5-4B · Qwen3.5-9B

✨ More intelligence, less compute.
These small models are built on the same Qwen3.5 foundation — native multimodal, improved architecture, scaled RL:
• 0.8B / 2B → tiny, fast, great for edge device
• 4B → a surprisingly strong multimodal base for lightweight agents
• 9B → compact, but already closing the gap with much larger models
And yes — we’re also releasing the Base models as well.
We hope this better supports research, experimentation, and real-world industrial innovation.
Hugging Face: https://t.co/wFMdX5pDjU
ModelScope: https://t.co/9NGXcIdCWI

[Minty @MintyNFT]

@AIAdsApps App

[Minty @MintyNFT]

@oliviscusAI Not this one. We need a free opensource one. Better memory and context mgmt.

[Minty @MintyNFT]

@sudoingX Damn great find. Gonna see if I can implement tonight after work.

[Minty @MintyNFT]

@erickpinos @PromptLLM Ya check it out. A lot of people forked their own and launched copies hahaha

[Minty @MintyNFT]

@gregschoeninger I’m getting about 35tps on 36gb mbp m3. Damn slow

[Minty @MintyNFT]

@BrianRoemmele @grok So it’s not good? About to set it up tonight and test lol

[Minty @MintyNFT]

Damn gonna setup a new bot with ollama Qwen 3.5 35B 4bit tonight and test it out.

Hopefully it’s as good as all the tweets im seeing. 😳

[OpenClaw🦞 @openclaw]

OpenClaw 2026.2.25 🦞

💓 Heartbeat DM delivery restored
🤖 Subagent delivery overhaul
🧵 Slack thread session fixes
🛡️ Reaction auth across all channels
🔒 Security hardening

More bars than Kendrick's cousin's changelog 🎤https://t.co/NkPtXjX4Ro