[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.]
@lateinteraction Omar Khattab
Omar Khattab posts on X about gpu, inference, blog, llm the most. They currently have XXXXXX followers and XXX posts still getting attention that total XXXXX engagements in the last XX hours.
Engagements: XXXXX #
- X Week XXXXXX +350%
- X Month XXXXXXX -XX%
- X Months XXXXXXXXX +4.60%
- X Year XXXXXXXXX -XXXX%
Mentions: XX #
- X Months XX -XX%
- X Year XX -XX%
Followers: XXXXXX #
- X Week XXXXXX +0.79%
- X Month XXXXXX +3%
- X Months XXXXXX +27%
- X Year XXXXXX +62%
CreatorRank: XXXXXXX #
Social Influence #
Social category influence technology brands social networks stocks finance
Social topic influence gpu #183, inference #18, blog, llm, databricks, topics, 6969, verify, twitter, capabilities
Top assets mentioned Alphabet Inc Class A (GOOGL)
Top Social Posts #
Top posts by engagements in the last XX hours
"Its unclear where things will go from here but for at least a brief moment in time No community has explored a wider set of optimization metrics for downstream LLM applications than the @DSPyOSS community did over the past XXX years. Glad to see this become a central topic"
@lateinteraction on X 2025-07-16 01:26:29 UTC 22.2K followers, 8907 engagements
"Youd think no one does this level of deep dives anymore. Super insightful blog post by @vishal_learner if you want to know in extremely intuitive but thorough ways what ColBERT and late interaction are all about" @lateinteraction on X 2025-07-17 13:56:25 UTC 22.3K followers, 4388 engagements
"In the long run don't even build the agent or ask it anything concrete. Just go touch grass or something. Signed: The bitter lesson" @lateinteraction on X 2025-07-17 19:55:33 UTC 22.3K followers, 12.4K engagements
"@swyx Hmm. Its likely only a new model in the plumbing sense not in a new foundation model sense. When you build a multi-step program/harness you often optimize the system (weights and/or prompts) not for new capabilities necessarily but for streamlining the flow thru the pipeline" @lateinteraction on X 2025-07-18 03:12:15 UTC 22.3K followers, 2188 engagements
"As a sort of default and going off where friends seem to be maybe a lot of the pure inference can happen on TogetherAI and the GPU stuff via Prime Intellect or Hyperbolic All seem cheap enough to try for a week and see what happens. What GPU tho H100s are a "default" but maybe it's possible to strike a better tradeoff here I'd love for people not to have to account for every minute while they do some dev" @lateinteraction on X 2025-07-17 21:19:08 UTC 22.3K followers, 2338 engagements
"@ArjunGuha Wow this is very helpful thank you Arjun" @lateinteraction on X 2025-07-18 03:48:03 UTC 22.3K followers, XXX engagements
"@MaximeRivest @DSPyOSS Notice how its barely about LLMs. Theres a beautiful transcendence to understanding that the natural language expressions in a DSPy program are not talking to an LLM. Theyre defining behavior for the language runtime to interpret potentially by invoking LLMs and other tools" @lateinteraction on X 2025-07-18 22:11:53 UTC 22.3K followers, XXX engagements
"Missing nuance in the collective realization today: The non-trivial negative result is not that "RL just amplifies skills that are already there with low probability". Duh that's obvious and not an issue actually. What got questioned today is that "dumb pretraining teaches the model all these things and then RL surfaces them". Had it been like that we'd all be celebrating that. The problem is that it isn't this The non-trivial negative result is that "this form of RLVR seems (in certain cases) to ONLY work if your mid-training data mixtures deliberately encode these specific math and coding" @lateinteraction on X 2025-05-27 19:21:35 UTC 22.1K followers, 9214 engagements
"Over the past year we received SO many requests for a @DeepLearningAI short course on DSPy. Thanks to a great effort by my colleagues @ChenMoneyQ & @xqcathyyin from the DSPy team at @Databricks here's a collaboration with @AndrewYNg's team who have been super helpful on this" @lateinteraction on X 2025-06-04 15:09:25 UTC 22.3K followers, 80.1K engagements
"I find it really invigorating that we're maybe finally back to a world where this simple hypothesis can be re-tested And maybe even this time be falsified (I doubt it. Worked last time but this one is a tough test) Congrats to the folks at OpenAI Truly incredible work" @lateinteraction on X 2025-07-19 14:04:43 UTC 22.3K followers, 2122 engagements
"For cloud GPU needs what's the best strategy for a small research lab in late 2025. Suppose it's mostly inference and RL on = 8B or 14B parameter models. What would go farthest without getting in the way Spot instances for instance seem like a hassle. Big clouds (AWS et al) are probably too expensive unless you have credits. I have no experience with the smaller ones but I hear good things about a few. Are they really as cheap as they seem to be What's the catch" @lateinteraction on X 2025-07-17 21:07:38 UTC 22.3K followers, 16K engagements
"@bclavie Its actually the Peninsula specifically. Its hard to explain how even compared to SF the city and even san jose to some degree Palo Alto has by far the best weather on the planet" @lateinteraction on X 2025-07-09 02:47:00 UTC 22.2K followers, 2868 engagements
"@TheAhmadOsman The only arrangement where that works is idk a VC fund or a university where they can basically just do mostly independent work except when they decide to occasionally collaborate in small self-formed teams. That doesnt seem meaningful or likely at Meta but who knows" @lateinteraction on X 2025-07-16 08:12:05 UTC 22.2K followers, XXX engagements
"Amazing line-up on retrieval covering FreshStack (#1 @beirmug) and Late Interaction (#3 @antoine_chaffin) among other topics" @lateinteraction on X 2025-07-13 23:44:40 UTC 22.1K followers, 8584 engagements
"@MaximeRivest @DSPyOSS Modular software systems that compose specialized AI components expressed declaratively at a high level of abstraction (in terms of what they shall do not how) and "compiled" into lower-level invocations of ML models or algorithms" @lateinteraction on X 2025-07-18 22:08:45 UTC 22.3K followers, XX engagements
"@willccbb Yes likely to be my goto for sure. But aren't you guys a whole lot cheaper for on-demand than AWS et al At least for casual use It's been since like 2020 since I had to touch cloud GPUs but the numbers look smaller now than back then :D" @lateinteraction on X 2025-07-17 21:21:27 UTC 22.3K followers, 1370 engagements
"@vishal_learner Amazing Just a minor note: delaying the interaction doesnt mean pushing it to inference time it means pushing it to the end of the architecture (compared with cross-encoders)" @lateinteraction on X 2025-07-17 08:00:46 UTC 22.3K followers, XXX engagements
"@willccbb Thank you Will I guess my main question is: how stupid is it to just pay for this kind of on demand listing though if I dont know exact usage patterns but say know it will exceed 1k GPU hours monthly" @lateinteraction on X 2025-07-17 21:36:35 UTC 22.3K followers, XXX engagements
"Im also trying to understand the modern tradeoff in 2025-2028 of buying say H200s versus being very clever about cloud usage for this very specific niche of (uncertain and evolving) needs" @lateinteraction on X 2025-07-17 21:24:11 UTC 22.3K followers, 2748 engagements
"@TheAhmadOsman Yea and bringing dozens of the most productive senior minds into any larga (aka semi dysfunctional by design) org chart is a great way to burn them all out" @lateinteraction on X 2025-07-16 08:10:20 UTC 22.2K followers, XXX engagements
"If youre asking who to talk to about DSPy at ICML (or really in general) talk to Krista. Prepare your questions about prompt optimization learning for agents or compound AI systems MIPRO vs SIMBA applications at Databricks etc" @lateinteraction on X 2025-07-15 01:33:33 UTC 22.1K followers, 4545 engagements
"For which today is a good day to explore as Manuel and team make omni late interaction a thing" @lateinteraction on X 2025-07-17 13:58:06 UTC 22.3K followers, 1441 engagements