[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.] [@neural_avb](/creator/twitter/neural_avb) "What platform do yall use to spin up a remote GPU server On the market looking for one"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1945714656372498731) 2025-07-17 05:18:18 UTC 2154 followers, 1652 engagements "The main challenge is probably partial observability: the real factors that affect stock prices is vast your agent will likely see a partial noisy incomplete state. So observation space needs to be detailed enough to allow the network to make decisions. Just past trends probably not enough"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1946256280344985725) 2025-07-18 17:10:31 UTC 2154 followers, XXX engagements "Generally all projects/tutorials I do for YT are available only on my Patreon. This is the main way my channel makes revenue. For the context engineering / dspy tutorial though I have decided to make it open-source. I'll make the repo public after the video drops. Patrons will still get exclusive access to slides write-ups and I will prepare some additional code materials for yall"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1947677335982862638) 2025-07-22 15:17:17 UTC 2155 followers, 1188 engagements "I miss the days where you could learn nasty little optimization tricks and find cool algorithmic ideas by reading NLP research papers from the top labs. These "technical reports" are more about withholding information than providing them"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1944803074633293884) 2025-07-14 16:56:00 UTC 2149 followers, XXX engagements "@JacksonAtkinsX Kinda interesting. I feel like LLMs can differ in capabilities so much depending on size provider training data etc - tuning the prompts to just one LLM may not always guarantee good performance on other models"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1946353334857646427) 2025-07-18 23:36:11 UTC 2138 followers, XXX engagements "@Hesamation Mental note/Social Promise: I need to make a proper video on Context Engineering"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1945264165641228562) 2025-07-15 23:28:13 UTC 2141 followers, 2697 engagements "Do you think AGI can be achieved through back propagation and gradient descent"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1946423081733411156) 2025-07-19 04:13:20 UTC 2141 followers, 1400 engagements "Here are X must-know RL libraries for single-agent and multi-agent environments. I am using most of these right now to make my multi-agent collaborative navigation environment - so it's a great time to post this list"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1944289959126380972) 2025-07-13 06:57:04 UTC 2154 followers, XXX engagements "Today I started ideating a dedicated video on Context Engineering. I'll cover the main approaches some research trends and a walkthrough of the concepts with DSPy. This git repo will be my muse for this one"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1945332815337238734) 2025-07-16 04:01:00 UTC 2155 followers, 17.1K engagements "The best dspy tutorial is actually shipped with dspy. It is the inspect_history function. Tells you everything you need to know about how your signatures are converted into system/user prompts. And how different modules do multi-step/single-step LM calls"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1947698525535670387) 2025-07-22 16:41:29 UTC 2154 followers, 2238 engagements "My Reasoning with GRPO tutorial has 28K+ views in XX days This is the highest number I have gotten from a blogpost. π«‘ It explains the concepts visually starting from foundational concepts. Includes code snippets and results from my own experiments. Link in comment"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1946246614747619741) 2025-07-18 16:32:07 UTC 2154 followers, 14.8K engagements "I think I am cooking the Context Engineering with DSPy tutorial hard. The blueprint. ππ½"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1945532666926846333) 2025-07-16 17:15:08 UTC 2141 followers, 12.9K engagements "I have started making slides and illustrations for the Context Engineering tutorial My goal is to upload by Thursday. π€π½ Btw I already have a DSPy tutorial and a pretty comprehensive theoretical overview of RAGs. They are optional pre-requisites"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1947330319389962574) 2025-07-21 16:18:22 UTC 2155 followers, 8015 engagements "Halfway through the Context Engineering w DSPy tutorial. I quite like where it's going. I am taking the simple but classic "write a joke" prompt and creating a full 35-ish minute journey out of it - going through each concept one by one"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1946083498684457430) 2025-07-18 05:43:57 UTC 2154 followers, 4170 engagements "Today I looked deeper into dspy.Refine (& BestOfN). Its insane just how much heavy lifting these modules can do. Define a reward function for any dspy module and Refine will generate responses optimized for to maximize the reward. The way it does so is very interesting. (thread)"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1945616221648339295) 2025-07-16 22:47:09 UTC 2154 followers, 15.4K engagements "@willccbb The ancient art of nomenclature"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1947671283182764195) 2025-07-22 14:53:14 UTC 2155 followers, XXX engagements "This taxonomy for Context Engineering is sooo handy. I'm making a comprehensive video tutorial on this using DSPy and my goal is to incorporate atleast one major algorithm from all the blocks"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1946288694882685317) 2025-07-18 19:19:19 UTC 2154 followers, 72.1K engagements "3 things I am looking forward in LLM Reasoning research: X. Applying RL on open-ended tasks instead of just strictly verifiable tasks X. A self-supervised way to assign credits to individual/group of tokens given a response X. Advantage vanishing issues as model gets smart"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1946894354653610327) 2025-07-20 11:26:00 UTC 2133 followers, XXX engagements "This new @lossfunk article is full of some of the wackiest practical tips for context engineering. Very interesting read"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1947690901091823649) 2025-07-22 16:11:11 UTC 2154 followers, 1539 engagements "Context Engineering tutorial is in it's last legs I thought I'll have a section on Python sandboxes and autonomous coding but imma reduce it to a footnote instead. Evaluation is kinda important so it's taking priority. I am scared how long this tutorial is gonna be"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1946810490488774674) 2025-07-20 05:52:45 UTC 2155 followers, XXX engagements "Context Engineering will run 1+ hours. Lot of editing for me the next X days.π Good educational content need to repeat key concepts to reinforce the viewer's mind. And engaging YT videos need to be tight and reduce fat. This mental doublethink is so hard to balance"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1947421429592359250) 2025-07-21 22:20:24 UTC 2155 followers, XXX engagements "Pretty nice article about different ways long context can actually harm generation quality"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1947193605656867016) 2025-07-21 07:15:07 UTC 2155 followers, XXX engagements "@mickel_liu Gotta love how classic RL paradigms are now being directly applied to LLM training. Cool poster"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1946101360354570657) 2025-07-18 06:54:55 UTC 2155 followers, XX engagements "I have a Deep Learning YouTube channel with 20K+ subs 500K views. I have a Patreon with over XX active members over XXX lifetime members. I also do a full time job side projects and occasional freelancing. Ask me anything Ill reply to everyone"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1946598696982130935) 2025-07-19 15:51:10 UTC 2138 followers, 12.5K engagements "This started as a Context Engineering tutorial. I feel it's borderline a course now Anyway I'll be shooting in a few hours"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1946908447359791492) 2025-07-20 12:22:00 UTC 2155 followers, 3935 engagements "Dear algorithm show this only to people interested in Reinforcement Learning"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1944784369425805625) 2025-07-14 15:41:40 UTC 2155 followers, 47.5K engagements "A dspy module to do multi-hop HyDE search over a database of jokes. Generate a hypothetical doc from user query - hybrid search + rrf - generate another doc from fetched results - hybrid search + rrf - repeat. The building block of RAG is just good agentic searchππ½"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1946695345678741717) 2025-07-19 22:15:12 UTC 2155 followers, 5150 engagements "So the field of memory management for context engineering goes way deeper than I thought. I'll cover the greatest hits in my upcoming tutorial but this literally needs a whole week of research to do justice"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1946289886773870691) 2025-07-18 19:24:03 UTC 2141 followers, XXX engagements "Repo: Shoutout to @Hesamation for posting it"  [@neural_avb](/creator/x/neural_avb) on [X](/post/tweet/1945340294662078766) 2025-07-16 04:30:43 UTC 2141 followers, XXX engagements
[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.]
@neural_avb
"What platform do yall use to spin up a remote GPU server On the market looking for one" @neural_avb on X 2025-07-17 05:18:18 UTC 2154 followers, 1652 engagements
"The main challenge is probably partial observability: the real factors that affect stock prices is vast your agent will likely see a partial noisy incomplete state. So observation space needs to be detailed enough to allow the network to make decisions. Just past trends probably not enough" @neural_avb on X 2025-07-18 17:10:31 UTC 2154 followers, XXX engagements
"Generally all projects/tutorials I do for YT are available only on my Patreon. This is the main way my channel makes revenue. For the context engineering / dspy tutorial though I have decided to make it open-source. I'll make the repo public after the video drops. Patrons will still get exclusive access to slides write-ups and I will prepare some additional code materials for yall" @neural_avb on X 2025-07-22 15:17:17 UTC 2155 followers, 1188 engagements
"I miss the days where you could learn nasty little optimization tricks and find cool algorithmic ideas by reading NLP research papers from the top labs. These "technical reports" are more about withholding information than providing them" @neural_avb on X 2025-07-14 16:56:00 UTC 2149 followers, XXX engagements
"@JacksonAtkinsX Kinda interesting. I feel like LLMs can differ in capabilities so much depending on size provider training data etc - tuning the prompts to just one LLM may not always guarantee good performance on other models" @neural_avb on X 2025-07-18 23:36:11 UTC 2138 followers, XXX engagements
"@Hesamation Mental note/Social Promise: I need to make a proper video on Context Engineering" @neural_avb on X 2025-07-15 23:28:13 UTC 2141 followers, 2697 engagements
"Do you think AGI can be achieved through back propagation and gradient descent" @neural_avb on X 2025-07-19 04:13:20 UTC 2141 followers, 1400 engagements
"Here are X must-know RL libraries for single-agent and multi-agent environments. I am using most of these right now to make my multi-agent collaborative navigation environment - so it's a great time to post this list" @neural_avb on X 2025-07-13 06:57:04 UTC 2154 followers, XXX engagements
"Today I started ideating a dedicated video on Context Engineering. I'll cover the main approaches some research trends and a walkthrough of the concepts with DSPy. This git repo will be my muse for this one" @neural_avb on X 2025-07-16 04:01:00 UTC 2155 followers, 17.1K engagements
"The best dspy tutorial is actually shipped with dspy. It is the inspect_history function. Tells you everything you need to know about how your signatures are converted into system/user prompts. And how different modules do multi-step/single-step LM calls" @neural_avb on X 2025-07-22 16:41:29 UTC 2154 followers, 2238 engagements
"My Reasoning with GRPO tutorial has 28K+ views in XX days This is the highest number I have gotten from a blogpost. π«‘ It explains the concepts visually starting from foundational concepts. Includes code snippets and results from my own experiments. Link in comment" @neural_avb on X 2025-07-18 16:32:07 UTC 2154 followers, 14.8K engagements
"I think I am cooking the Context Engineering with DSPy tutorial hard. The blueprint. ππ½" @neural_avb on X 2025-07-16 17:15:08 UTC 2141 followers, 12.9K engagements
"I have started making slides and illustrations for the Context Engineering tutorial My goal is to upload by Thursday. π€π½ Btw I already have a DSPy tutorial and a pretty comprehensive theoretical overview of RAGs. They are optional pre-requisites" @neural_avb on X 2025-07-21 16:18:22 UTC 2155 followers, 8015 engagements
"Halfway through the Context Engineering w DSPy tutorial. I quite like where it's going. I am taking the simple but classic "write a joke" prompt and creating a full 35-ish minute journey out of it - going through each concept one by one" @neural_avb on X 2025-07-18 05:43:57 UTC 2154 followers, 4170 engagements
"Today I looked deeper into dspy.Refine (& BestOfN). Its insane just how much heavy lifting these modules can do. Define a reward function for any dspy module and Refine will generate responses optimized for to maximize the reward. The way it does so is very interesting. (thread)" @neural_avb on X 2025-07-16 22:47:09 UTC 2154 followers, 15.4K engagements
"@willccbb The ancient art of nomenclature" @neural_avb on X 2025-07-22 14:53:14 UTC 2155 followers, XXX engagements
"This taxonomy for Context Engineering is sooo handy. I'm making a comprehensive video tutorial on this using DSPy and my goal is to incorporate atleast one major algorithm from all the blocks" @neural_avb on X 2025-07-18 19:19:19 UTC 2154 followers, 72.1K engagements
"3 things I am looking forward in LLM Reasoning research: X. Applying RL on open-ended tasks instead of just strictly verifiable tasks X. A self-supervised way to assign credits to individual/group of tokens given a response X. Advantage vanishing issues as model gets smart" @neural_avb on X 2025-07-20 11:26:00 UTC 2133 followers, XXX engagements
"This new @lossfunk article is full of some of the wackiest practical tips for context engineering. Very interesting read" @neural_avb on X 2025-07-22 16:11:11 UTC 2154 followers, 1539 engagements
"Context Engineering tutorial is in it's last legs I thought I'll have a section on Python sandboxes and autonomous coding but imma reduce it to a footnote instead. Evaluation is kinda important so it's taking priority. I am scared how long this tutorial is gonna be" @neural_avb on X 2025-07-20 05:52:45 UTC 2155 followers, XXX engagements
"Context Engineering will run 1+ hours. Lot of editing for me the next X days.π Good educational content need to repeat key concepts to reinforce the viewer's mind. And engaging YT videos need to be tight and reduce fat. This mental doublethink is so hard to balance" @neural_avb on X 2025-07-21 22:20:24 UTC 2155 followers, XXX engagements
"Pretty nice article about different ways long context can actually harm generation quality" @neural_avb on X 2025-07-21 07:15:07 UTC 2155 followers, XXX engagements
"@mickel_liu Gotta love how classic RL paradigms are now being directly applied to LLM training. Cool poster" @neural_avb on X 2025-07-18 06:54:55 UTC 2155 followers, XX engagements
"I have a Deep Learning YouTube channel with 20K+ subs 500K views. I have a Patreon with over XX active members over XXX lifetime members. I also do a full time job side projects and occasional freelancing. Ask me anything Ill reply to everyone" @neural_avb on X 2025-07-19 15:51:10 UTC 2138 followers, 12.5K engagements
"This started as a Context Engineering tutorial. I feel it's borderline a course now Anyway I'll be shooting in a few hours" @neural_avb on X 2025-07-20 12:22:00 UTC 2155 followers, 3935 engagements
"Dear algorithm show this only to people interested in Reinforcement Learning" @neural_avb on X 2025-07-14 15:41:40 UTC 2155 followers, 47.5K engagements
"A dspy module to do multi-hop HyDE search over a database of jokes. Generate a hypothetical doc from user query - hybrid search + rrf - generate another doc from fetched results - hybrid search + rrf - repeat. The building block of RAG is just good agentic searchππ½" @neural_avb on X 2025-07-19 22:15:12 UTC 2155 followers, 5150 engagements
"So the field of memory management for context engineering goes way deeper than I thought. I'll cover the greatest hits in my upcoming tutorial but this literally needs a whole week of research to do justice" @neural_avb on X 2025-07-18 19:24:03 UTC 2141 followers, XXX engagements
"Repo: Shoutout to @Hesamation for posting it" @neural_avb on X 2025-07-16 04:30:43 UTC 2141 followers, XXX engagements
/creator/twitter::1754194661084983296/posts