steadyconfidence high
pump.fun 13DexScreener 3Social 9
Growth · per hour, 72hobserved
pump.funNew poolsDexScreenerSocial
LLM
Symbol since 23h ago · updated 5s ago
Readingscalculated
attention63
replication+2311 distinct token names across 16 tokens in 24h
velocity+94 mentions in last 6h against 28 in prior 24h, −43%
creators+1122 distinct creator wallets, accounts and threads in 24h
spread+116 channels across 3 source families in 24h
market+10$283.7k traded across 10 matching tokens in 24h
weights · replication 30 · velocity 20 · creators 20 · spread 15 · market 15. a reading still waiting is left out and weights rescale.
Tokens carrying it · 72hobserved
| Current | Current · robinhood | $137.3k | |
| Inferrr | INFERRR · solana | $73.2k | |
| 1claw AI | 1clawAI · base | $36.9k | |
| Kimi K3 | Kimi · solana | $32.8k | |
| Accred | Accred · solana | $1.3k | |
| Global X LLM ETF | LLMA · solana | $802 | |
| tavily.com | TAVILY · solana | $667 | |
| Accred.sh | Accred · solana | $567 |
13 more carry this name
Related strains · 72hcalculated
accred7 shared
breaks3 shared
chanky llm2 shared
chanky2 shared
'PHYSICAL8 shared'SEARCH3 shared'KEY2 shared'GROUND2 sharedPosts and threadsobserved
16:15:02post Spark X2.5: on-device agentic AI with native 1M token context. No chunking, no cloud dependency, no lost context. Runs on vLLM, Ollama, llama.cpp. Apache 2.0.
Full deployment guide: aiadoptionagency.com/spark-x2-5-o...23:33:42post though maybe a diffusion LLM with temporal tokens is similar23:09:05post An LLM is an autoregressive model that repeatedly computes a probability distribution over the next token given the current context. Once a token is emitted, it becomes part of that context and isn’t internally revised. Subsequent tokens are generated conditio…22:49:06post I'm pretty sure I'm sending some hostility here. Okay, given my general dislike of AI, I think I've burned enough of the federal government's LLM tokens.22:48:44post Yes, essentially the LLM predicts a single token based on the whole 'context', the context is your prompt, some instructions and *all of the previous output*, and the LLM has no way to know the difference. It just predicts what token come next, then the orches…22:46:02post A thing I learned not too long ago is that when an LLM selects a token that leads it down an incorrect path (factually), it has no way to back up, or even know that it should back up, it's just on to the next likely bit. Literally every LLM is just posting the…22:34:22post GPT-6.1 Sol costs one-fifth of GPT-6 Astra on token pricing while delivering nearly identical performance on coding and professional work. #openai #llm
https://bymachine.news/openai-devday-2026-gpt-6-1-sol-dots-codex21:11:00post Yeah -- if you're selling a system that you call an LLM and you're sick of being embarrassed by the fact that next-token prediction is a terrible way to do arithmetic, you run a little text classifier up front and shunt math problems to an actual calculator.18:05:19post SHIPPED: OpenAI launched office suite features within ChatGPT, competing directly with traditional productivity software.
#AI #LLM17:22:13post My favorite part of interacting with LLM is spending tokens to train it to stop talking like a right-wing influencer.
"Battle" "Cutting" "Slicing" "Breaking" ...Yeah great, writing something to convert a stream into another stream is a battle for REAL MEN; S…
Searches and readingobserved
Not in search trends or reading lists.
Research · 'LLM
A ten part report written only from evidence on this sheet, with every claim linked to its source.