[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]

@snipeship "@_lukaemon might be a bit more complex than that but it's rumored to be best-of-n / synthesis. the model is still o3"
@snipeship on X 2025-07-27 02:59:32 UTC XXX followers, XX engagements

"@canoozie Beautiful How is perf like with the 4bit quant"
@snipeship on X 2025-07-28 22:37:54 UTC XXX followers, XX engagements

"@pvncher @thdxr Opus w/ reasoning is the SOTA coding agent as well. Doesn't think for too long so it's great in terms of speed. Sonnet will still be a better agent with reasoning tokens. Pair it up with the smartest reasoners (o3/o3-pro) and look at the best results you've ever seen with LLMs"
@snipeship on X 2025-07-13 01:42:53 UTC XXX followers, XXX engagements

"@mckaywrigley I'd recommend @RepoPrompt for this. Also build a plan using o3-pro. Context Builder + MCP will: - Reduce onboarding time - Increase reliability - Save lots of tokens It really is the best Context Engineering tool"
@snipeship on X 2025-07-23 03:15:39 UTC XXX followers, 6000 engagements

"The ultimate truth: July 2025 LLM coding meta - o3-pro for research/planning - o3-pro for code review - o3-pro for debugging - opus X for agent tool calling driver - o3-pro for first pass code implementation - opus X for iterative changes to existing code - Gemini XXX Pro rarely for long context Opus on CC + o3-pro on @RepoPrompt's MCP. This is SOTA & it's not even close"
@snipeship on X 2025-07-24 00:46:18 UTC XXX followers, 4701 engagements

"@jfFxjfgLrmKRr @ericzakariasson Debugging hallucinated code because you cheaped out on the model is more expensive"
@snipeship on X 2025-07-23 22:06:17 UTC XXX followers, XXX engagements

"We just hit 100M today Special shoutout to @bunjavascript (we're running entirely on Bun)@claude_code & @RepoPrompt (we would not be shipping as fast without them) It's been a wild ride and we're just getting started"
@snipeship on X 2025-07-28 17:19:09 UTC XXX followers, 2205 engagements

"@GosuCoder Nvidia building something like GeForce Now for inference would be nice call it CUDA Now lol"
@snipeship on X 2025-07-27 05:13:30 UTC XXX followers, XXX engagements

"This is a great launch. Remember always: X. export MAX_THINKING_TOKENS=39999 X. claude config set --global parallelTasksCount X X. /model opus & call it a day"
@snipeship on X 2025-07-24 22:11:33 UTC XXX followers, XXX engagements

"Question: Why is the default when adding a new @Cloudflare site "Block AI training bots" & "Instruct AI bots to not scrape content" Shouldn't it be the other way round (show the option but not have it on by default) At first glance (imo) it might be a disadvantage to not be in the training set unless you're dealing with copyrighted data etc. - but I don't think the majority of websites land in that category. cc @dok2001"
@snipeship on X 2025-07-27 12:24:08 UTC XXX followers, XXX engagements