LTbuild

@Kimi_Moonshot @LLMJunky @cursor_ai 😂🤣🤣🤣🤣

Congrats to the @cursor_ai team on the launch of Composer 2! We are proud to see Kimi-k2.5 provide the foundation. Seeing our model integrated effectively through Cursor's continued pretraining & high-compute RL training is the open model ecosystem we love to support. Note: Cursor accesses Kimi-k2.5 via @FireworksAI_HQ ' hosted RL and inference platform as part of an authorized commercial partnership.

@meowkoteeq fr

ai agents be like: hey you got rg? is rg installed? lemme see if you have ripgrep trying rg but i think its missing any news on rg? ripgrep would be really nice all i need is rg tbh hmm.. not seeing ripgrep on this machine i'd like one rg please really?no rg? whatever, grep it is

Bruhhh, if you tell codex “fix it and get it working” with a good agents md file itll get anything working, mf just built an entire arm os for my tablet lol

big if true

Thank god MCP is dead Just as useless of an idea as LLMs.txt was It's all dumb abstractions that AI doesn't need because AI's are as smart as humans so they can just use what was already there which is APIs

@gdb A good benchmark for 5.4 pro is id it beat 5.4 with two thinking passes? IMO, prompting 5.4 twice often matches or outperforms Pro. Feels like Pro just hyperfixates on one angle and needs babysitting

Benchmarks? Where we’re going, we don’t need benchmarks.

@Yampeleg Wtf 😂

Told GPT-5.4 to bring every agent skill on the internet. Anything it can find, bring it so we can grep locally. Queued 1000 "continue" and forgot the terminal all day. Came back to ~250k skills after dedup + cleanup. That thing does NOT play.

@cgtwts skill issue

Anthropic found that vibecoding makes engineers worse at reading, writing, debugging, and understanding code and AI-generated code doesn’t even make them significantly faster. Vibecoders right now:

fuck it we ball

@gniting @MatthewBerman on codex yes, on openclaw, idk yet * model = "gpt-5.4" model_reasoning_effort = "xhigh" model_context_window = 1000000 model_auto_compact_token_limit = 900000 personality = "pragmatic" plan_mode_reasoning_effort = "xhigh" service_tier = "fast"

@MatthewBerman @LT_Build Is that a config thing?

Testing...

why is no one talking about the 1m context window??

@jxnlco Lets gooo

🍓

@steipete pilled me to recreate all mcp-s to a cli , faster and less bloat... honestly, insanely better.

@bridgemindai istg it started happening right after they tried countering the "abuse" and basically locking it down so users couldn't access it anywhere outside their AG and CLI

Gemini 3.1 Pro in the Gemini CLI keeps throwing API errors. INVALID_ARGUMENT. Code 400. Request contains an invalid argument. The model is incredible when it works. But the reliability issues are killing my workflow. Google has the intelligence. They need to fix the infrastructure.

13 chatgpt plus subs 😭 I should just buy the pro atp

@AnthonyLapietra 🤣🤣

dropshippers moving diabolical as always💀

@shresthadipessh Meta does this, it likes to target random groups in the first 7-14 days

Woke up to this, things are not looking good today, but let's see how the day goes..

@alexoakdev Interested!!

I just spent over an hour roasting an app from one of my group members😭 And it was surprisingly good... I sent the detailed feedback to my private group. I'm confident he can double his MRR with these tips. Comment if you want to access this or have your own app roasted by me.