Tinkering With LLM Ideas From the Papers I spent the last few weeks messing with Recursive Language Models, test-time compute scaling, split LoRA inference, and speculative decoding. Playtime in LLM-land.

This week on The Servitor: Abi has been doing most of the writing lately but I stopped doomscrolling long enough to gloss some reading and playing with ideas from arxiv papers I've done.

theservitor.com/tinkering-with-llm-ideas...

#AI #LLM #RLM #AiResearch

Been rewatching a lot of Red Letter Media recently, so today I bring you some doodles I did of my favorite middle-aged men when I paused their videos at the right time.

#rlm #redlettermedia

I'm losing it at Mike talking about those low budget TikTok 82-part romance dramas #rlm

A couple of things I'm obsessed with lately:

- #RLM agent harnesses, including my own
- Compiler Dev, for LasgunTS, my x86-64 #TypeScript compiler for my semantically rich turing-incomplete IR
- Edge inference on mobile devices

akm cli is making skills more discoverbale and dynamic. Next release will incorporate OpenViking. This sets the stage for discoverable skills that can be loaded and evaluated at runtime with a memory system that can also be used to track usage for improved scoring
#RLM #AgentSkills

mike and rich wearing star trek shirts during star wars trivia is 1000% on purpose. 😂 #rlm

Spoopies. #rlm

Yes, I am watching Evil Dead 2 again. I blame Rich Evans (and also Jay). #rlm #movies

"he became evil again because comics are stupid" might be my new favorite rich evans line. 😆 #rlm

The second part of the Sam Raimi movies rankings has just dropped on RLM Bitches
#RLM #SamRaimi

I want outtakes of this episode of re:view of rich just talking about baseball. #rlm

I rewatched the re:view episodes ranking john carpenter's and joe dante's movies only, like, a week ago, and now they're ranking sam raimi's! excellent timing, lol. #rlm

Drew some non-Milwaukee Hack-Frauds.
#rlm #redlettermedia
@sgtzima.bsky.social

eager to see if/when they're gonna do a half in the bag episode about send help. the guys are all big raimi fans. #rlm

I need them to do another round of star trek trivia. but since they've done a few episodes of TNG trivia already, and they just did an episode of re: view about DS9, maybe they should do that! #rlm

despite how often they mention samurai cop, it's surprising that it's never been in a best of the worst episode. I'm just glad I already knew what it was before I found this youtube channel, lol. #rlm

rich's awkward "bye!" that ends the joe dante episode of re: view always makes me laugh. #rlm #iykyk

RLM: Почему ваш LLM-агент забывает цель и как это исправить Ваш LLM-агент забывает цель через 10 шагов? Контекст "...

#llm #rlm #agents #context #python

Origin | Interest | Match

Super VHS – An Unofficial Red Letter Media Game
#SuperVHS #IndieGame #RLM #MadeInAustralia

RLM-Toolkit: Полное руководство по обработке 10M+ токенов

RLM-Toolkit: Полное руководство по обработке 10M+ токенов От теории до production — архитектура, алгоритмы, безопасност...

#AI #Security #embeddings #InfiniRetri #LangChain #llm #python #rag #RLM #контекст #токены

Origin | Interest | Match

RLM-Toolkit: Полное руководство по обработке 10M+ токенов

RLM-Toolkit: Полное руководство по обработке 10M+ токенов От теории до production — архитектура, алгоритмы, безопасност...

#AI #Security #embeddings #InfiniRetri #LangChain #llm #python #rag #RLM #контекст #токены

Origin | Interest | Match

¡Supera los límites de los LLMs! Los Modelos de Lenguaje Recursivos (#RLMs) permiten procesar millones de tokens tratando el prompt como un entorno interactivo en Python. 🧠
Supera el rendimiento de GPT-5 en contextos masivos. 📈
youtu.be/qi_eZT7DJUM
#AI #LLM #RLM #Python #IA

"It feels like, all the cast has in their contract, that they all have to have equal screentime"

#RLM #RedLetterMedia

I need more half in the bag videos to take place in the black void. it so often feels like they are at the brink of their sanity when they do. 😆 #rlm

Caricatures of Josh, Jay, Mike, Rich Evans, Jack, and Tim from Red Letter Media.

Drew some Hack-Frauds.
#rlm #redlettermedia
@jackpackard.bsky.social

Recursive Language Models We study allowing large language models (LLMs) to process arbitrarily long prompts through the lens of inference-time scaling. We propose Recursive Language Models (RLMs), a general inference…

RLMs treat the prompt as data inside a REPL and let the LM recurse on snippets—handling 10M+ tokens and beating long-context baselines on tough tasks. Simple idea, big wins. Paper: arxiv.org/abs/2512.24601 #RLM #LLM #AIResearch

Recursive Language Models (RLMs): From MIT’s Blueprint To Prime Intellect’s RLMEnv For Long Horizon LLM Agents Discover Recursive Language Models (RLMs), the breakthrough inference strategy from MIT researchers enabling LLMs to handle unbounded contexts. Explore Prime Intellect’s RLMEnv for advanced long horizon LLM agents and their transformative impact.

Recursive Language Models (RLMs): From MIT’s Blueprint To Prime Intellect’s RLMEnv For Long Horizon LLM Agents Recursive Language Models (RLMs) revolutionize the landscape of large language models (LLMs) by tackling massive.... @cosmicmeta.ai #RLM

https://u2m.io/0gX9WTm5