[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.]

@casper_hansen_ Avatar @casper_hansen_ Casper Hansen

Casper Hansen posts on X about open ai, o3, token, moonshot the most. They currently have XXXXX followers and XXX posts still getting attention that total XXXXXX engagements in the last XX hours.

Engagements: XXXXXX #

Engagements Line Chart

Mentions: XX #

Mentions Line Chart

Followers: XXXXX #

Followers Line Chart

CreatorRank: XXXXXXX #

CreatorRank Line Chart

Social Influence #

Social category influence technology brands stocks

Social topic influence open ai, o3, token, moonshot, $googl #1193, banger, onboard, inference, domain, hybrid

Top assets mentioned Alphabet Inc Class A (GOOGL) Hundred Million (100M)

Top Social Posts #

Top posts by engagements in the last XX hours

"Remember MPT LLM-Foundry Megablocks Streaming Composer + fault-tolerant training All good things come to an end"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-17 17:08:16 UTC 8388 followers, 2921 engagements

"@kellogh @YouJiacheng X. Collect dataset of problem and solution X. Label if correct or incorrect using Lean or LLM-as-judge or third method X. Train verifier with labels Now you have parameterised the method"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-19 12:35:13 UTC 8395 followers, XX engagements

"@WolframRvnwlf I found this nugget yesterday. Surface-level this looks like everything you would want"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-21 12:47:23 UTC 8402 followers, XXX engagements

"@xlr8harder Find a hobby you enjoy. Read some books. Lift some weights"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-20 09:54:41 UTC 8402 followers, XXX engagements

"Step X of many: Last week I released a biomedical dataset of 521k samples. This week I released full-text embeddings (32k) with 2048 dimension from Qwen3 4B embedding model"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-09 19:47:16 UTC 8393 followers, 15.7K engagements

"Whos making an RL trainer but for spot instances Cheap RL is good RL. Spot goes down another one spins up"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-18 10:29:35 UTC 8393 followers, 1940 engagements

"The RL codebase I like the most: - The NanoGPT of RL - Supports multi-turn RL - Just 1k lines of code in Python - Data Tensor Sequence Parallel"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-17 09:45:32 UTC 8402 followers, 26.3K engagements

"Ever wanted to solve biomedicine Here you go thousands of applications can be created from this"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-17 11:16:56 UTC 8391 followers, 5982 engagements

"@Sebyverse Ever heard of RLHF or RLVR Thats how models like o3 R1 K2 etc. are trained. 80-90% of this training process is inference. So you need to build inference before you build training"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-21 13:13:27 UTC 8403 followers, XXX engagements

"@YouJiacheng Yes and the verifier can be trained on data that is verified in Lean or otherwise"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-19 12:24:08 UTC 8398 followers, 1410 engagements

"@TheZachMueller Looked it up gigabrain vibes on this one"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-21 12:42:45 UTC 8402 followers, XXX engagements

"@MaziyarPanahi @huggingface Im curious: why the choice to split on domain Couldnt you just train one good NER model with some token balancing"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-17 11:17:26 UTC 8387 followers, XXX engagements

"@HrishbhDalal I like Yann Lecuns ideas on this lately. The compounding token errors make sense. But I think we underestimate just how much more we can scale LLMs"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-19 12:04:53 UTC 8395 followers, XX engagements

"The OpenAI o3 alpha model might just be the open-source model that was waiting to finish because of the code contest"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-19 06:26:10 UTC 8398 followers, 1209 engagements

"@giffmana @Alibaba_Qwen Try to read the chat template and apply it in various ways over multiple turns"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-21 19:13:05 UTC 8402 followers, XXX engagements

"@giffmana @Alibaba_Qwen The Qwen3 hybrid chat template was a nightmare to manage in a multi-turn scenario. Apart from that a hybrid model should be as strong as the standalone version. If you cant do that then a routing system over a model harness is better"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-21 19:06:21 UTC 8402 followers, XXX engagements

"Why do we have no western DeepSeek or Moonshot Feels like the only close one would be Mistral but even they pale in comparison"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-17 16:19:43 UTC 8386 followers, 3623 engagements

"@HrishbhDalal Its easy to verify in the sense that you can just use LLM-as-a-judge but hard in the sense that you may not know how to measure success"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-19 11:58:10 UTC 8395 followers, XXX engagements

"thats a nice SOTA wanna join meta for $100m works every time"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-19 08:33:10 UTC 8398 followers, XXX engagements

"This is not a SMALL update. This is huge Give us this for every model please Qwen team🙏"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-21 17:32:55 UTC 8402 followers, 6132 engagements

"Google just deactivated my final ad-blocker that worked. So now its really time to move browser - what are people using these days"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-21 12:37:07 UTC 8402 followers, 2781 engagements

"Im seeing lots of people with the worst takes on IMO medals OpenAI and generally AI on the timeline. Where did we go wrong when we critique such a crazy achievement The craziest part that its just natural language"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-20 10:40:44 UTC 8402 followers, 1112 engagements

"@MaziyarPanahi has great food needs X different dressings to swallow it smh bro😂😭"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-18 10:05:57 UTC 8393 followers, XXX engagements

"We need something better than Nvidia or AMD if we want to scale up AGI and make it accessible. A simple 2x in compute performance every X years is not enough to satisfy where we are headed"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-21 14:11:36 UTC 8402 followers, XXX engagements

"This blog post is actually filled with golden nuggets on context engineering that is applicable to training agents with RL"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-19 07:05:05 UTC 8395 followers, 1018 engagements

"@QuixiAI @huggingface This can be annoying. It also seems unnecessary. I tried searching for an environment variable to change this but couldnt find one"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-18 15:50:03 UTC 8392 followers, XXX engagements

"@boneGPT Sam has a CS degree and previously coded"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-19 18:39:27 UTC 8402 followers, 2433 engagements

"@prashant_hq I didnt have the lite version"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-21 12:40:00 UTC 8402 followers, XXX engagements

"@colorado_plays @cursor_ai Im assuming you are on the max plan on Cursor I personally havent bumped into limits yet but I main o3. I heard Roo Code is a better Cline (fork but better)"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-18 10:37:35 UTC 8396 followers, XXX engagements

"@xlr8harder I would argue they already have surpassed the west. Look at number of researchers that are Chinese. Look at DeepSeek ByteDance Alibaba Moonshot and so many more producing SOTA models in all modalities"
@casper_hansen_ Avatar @casper_hansen_ on X 2025-07-13 08:18:48 UTC 8393 followers, XXX engagements