[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.]

@TheAhmadOsman Avatar @TheAhmadOsman Ahmad

Ahmad posts on X about gpu, open ai, agi, claude the most. They currently have XXXXXX 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: XXXXXX #

Followers Line Chart

CreatorRank: XXXXXXX #

CreatorRank Line Chart

Social Influence #


Social category influence technology brands XXXX% stocks #5845 finance XXXX% social networks XXXX% countries XXXX% celebrities XXXX%

Social topic influence gpu #43, open ai #213, agi #54, claude #64, money 0.85%, $googl #992, docs #33, all the 0.85%, debt #1104, bubble #887

Top accounts mentioned or mentioned by @najdollarsign @yacinemtb @rasbt @thekennwakanma @lateinteraction @wewillrd @mjyoke1111 @kuittinenpetri @startupmillyair @wreckedbymemes @eltokh7 @msuiche @qasimstatic @xeophon_ @dorialexander @redmondai @brundagecabins @itsquibloo @willccbb @fxdstudios

Top assets mentioned Alphabet Inc Class A (GOOGL)

Top Social Posts #


Top posts by engagements in the last XX hours

"nvidia market cap all pharma combined"
X Link @TheAhmadOsman 2025-10-03T23:36Z 25.7K followers, 6581 engagements

"I wanted a Mac M3 Ultra with 512GB RAM. Ahmad and Pewdiepie convinced me otherwise. this is a W in my book Buy a GPU"
X Link @TheAhmadOsman 2025-10-04T01:57Z 25.7K followers, 6423 engagements

"why don't i like ollama & what do i use on my AI server a thread of blogposts where i go over: my ai homelab setup what are inference engines how local llms actually work why i don't recommend ollama what do i use on my AI server the different use cases"
X Link @TheAhmadOsman 2025-09-03T14:22Z 25.9K followers, 225.9K engagements

"China saved opensource LLMs between July 16th and today these are the major releases: DeepSeek V3.2 GLM-4.6 (335B-A32B) Qwen3-VL-30B-A3B (Instruct & Thinking) Qwen3-VL-235B-A22B (Instruct & Thinking) Qwen3-Next 80B-A3B (Instruct & Thinking) GLM-4.5V (VLM 106B-A12B) DeepSeek V3.1 Doubao 1.6-Vision (multimodal tool-calling) Doubao Translation XXX (ByteDance XX Languages) ERNIE X1.1 (Baidu Reasoning) Hunyuan-MT-7B & Chimera-7B (Tencent Translation Specialists) MiniCPM-V XXX (8B) Tiny but GPT-4o-level VLM InternVL XXX (MASSIVE Multimodal Family of Models 1B to 241B Sizes) Step-3 (VLM 321B/38B)"
X Link @TheAhmadOsman 2025-10-07T04:59Z 25.9K followers, 159.8K engagements

"- be us - Larry & Sergey - at Stanford with a crawler and a dream - accidentally organize the entire internet - call it Google - build search email maps docs OS phones browser car satellite thermostat AI lab TPU farm and quantum computer - 2025 - everyone talking about AGI - OpenAI: we need data sensors feedback and scale - us: staring at Google Maps YouTube Gmail Android Waymo Pixel Fitbit Docs Calendar Street View and Earth Engine - "damn. guess we already did that." - YouTube: 2.6M videos/day - Android: 3B phones streaming sensor data 24/7 - Gmail: 1.8B inboxes of human priors - Search:"
X Link @TheAhmadOsman 2025-10-13T19:58Z 25.9K followers, 117.6K engagements

"been asked a lot today about the NVIDIA DGX Spark also getting lots of reactions on this old post of mine about it hoping to get my hands on one soon to review and expand a bit further on what it's good for and what not (plus it's pretty ngl want X on my desk haha) stay tuned"
X Link @TheAhmadOsman 2025-10-15T01:29Z 25.9K followers, 11K engagements

"DGX Spark - Ahmad's Analysis downgraded to no buy X the cost of a 5090 for XX% FP4 perf crippled bandwidth (273 GB/s worse than a MacBook Pro)"
X Link @TheAhmadOsman 2025-09-20T19:55Z 25.9K followers, 17.3K engagements

"- you are - a random CS grad with X clue how LLMs work - get tired of people gatekeeping with big words and tiny GPUs - decide to go full monk mode - X years later i can explain attention mechanisms at parties and ruin them - heres the forbidden knowledge map - top to bottom how LLMs actually work - start at the beginning - text tokens - tokens embeddings - you are now a floating point number in 4D space - vibe accordingly - positional embeddings: - absolute: i am position X - rotary (RoPE): i am a sine wave - alibi: i scale attention by distance like a hater - attention is all you need -"
X Link @TheAhmadOsman 2025-09-19T10:12Z 25.9K followers, 260.2K engagements

"be me Larry Ellison own a database empire a sailboat and a disdain for poor people OpenAI wants to build AGI needs compute a lot of compute like $300B GPU cluster in a volcano levels of compute "Hello. I own Oracle Cloud. Also Im rich." sign $300B GPU hosting deal with OpenAI nothing ships. no GPUs installed. no fans spinning. Oracle stock: skyrockets to the moon my net worth: +$100B GPUs still imaginary OpenAI raises a $X TRILLION round yes with a T investors lining up like it's a Taylor Swift concert I invest whered I get the money from the $300B deal I signed with them yes. OpenAI"
X Link @TheAhmadOsman 2025-09-13T00:18Z 25.9K followers, 1.8M engagements

"the top X most influential LLM releases that defined opensource AI LLaMA X Mistral 7B LLaMA X Qwen XXX DeepSeek R1"
X Link @TheAhmadOsman 2025-10-11T18:47Z 25.9K followers, 124K engagements

"- you are - a random CS grad with X clue how LLMs work - get tired of people gatekeeping with big words and tiny GPUs - decide to go full monk mode - X years later i can explain attention mechanisms at parties and ruin them - heres the forbidden knowledge map - top to bottom how LLMs actually work - start at the beginning - text tokens - tokens embeddings - you are now a floating point number in 4D space - vibe accordingly - positional embeddings: - absolute: i am position X - rotary (RoPE): i am a sine wave - alibi: i scale attention by distance like a hater - attention is all you need -"
X Link @TheAhmadOsman 2025-10-04T11:00Z 25.9K followers, 240.2K engagements

"Someday and that day will come youll be out of VRAM and desperate for compute. And on that day youll remember who told you first to Buy a GPU. The Godfather of TFLOPs was right"
X Link @TheAhmadOsman 2025-10-14T22:40Z 25.9K followers, 2620 engagements

"@WreckedByMemes LMStudio if your goal is easy ExLlamaV3 and then vLLM/Sglang"
X Link @TheAhmadOsman 2025-10-07T11:09Z 25.8K followers, 5023 engagements

"quick vram math for llms (weights only) fp16 = X bytes/param w4a16 = XXX bytes/param exl2 = XXXXX / XXX / XXXXX / XXXXXX bytes/param for XXX / XXX / XXX / XXX bpw rough vram per model size: 7b: fp16 14gb w4a16 3.5gb exl2 (4.0bpw) 3.5gb 13b: fp16 26gb w4a16 6.5gb exl2 (4.0bpw) 6.5gb 34b: fp16 68gb w4a16 17gb exl2 (4.0bpw) 17gb 70b: fp16 140gb w4a16 35gb exl2 (4.0bpw) 35gb 236b: fp16 472gb w4a16 118gb exl2 (4.0bpw) 118gb 405b: fp16 810gb w4a16 202gb exl2 (4.0bpw) 202gb do the math: vram = params bytes_per_param ex: 70b @ w4a16 XX XXX XXX = XX x XXX bytes XX gb fits on X 24gb gpus with room for"
X Link @TheAhmadOsman 2025-10-12T16:30Z 25.9K followers, 14.6K engagements

"i cancelled my Claude subscription and you should too. why 1.58-bit quantized models during daytime plus not getting opus X in claude code max plans limits cut in half X weeks ago no comms weekly limits without concrete numbers 5x/20x plans being actually 3x/8x of plus DMCA takedowns of repos that have to do with Claude Code windsurf no access to claude X cutting off openai api access "DEGRADED QUALITY" models incidents that they only acknowledged after being called and out without providing any further information support or refunds X years retention of all conversations and code all data"
X Link @TheAhmadOsman 2025-09-07T04:04Z 25.9K followers, 335.1K engagements

"men only want one thing (racks of NVIDIA DGX B300 GPUs)"
X Link @TheAhmadOsman 2025-10-11T06:35Z 25.9K followers, 27.9K engagements

"always be VRAM & TFLOPs maxxing anon Buy a GPU"
X Link @TheAhmadOsman 2025-10-14T23:34Z 25.9K followers, 1639 engagements

"some of anthropic rugpulls so far: X years retention of all conversations and code all data will be used for training 1.58-bit quantized models during daytime plus not getting opus X in claude code max plans limits cut in half X weeks ago no comms weekly limits without concrete numbers 5x/20x plans being actually 3x/8x of plus DMCA takedowns of repos that have to do with Claude Code one of which was my own personally windsurf no access to claude X cutting off openai api access what an absolutely horrible company"
X Link @TheAhmadOsman 2025-08-29T07:14Z 25.9K followers, 458.8K engagements

"@JamesLee1033176 multi-GPUs with Ollama is basically a waste of time compute and power that wrapper cannot do basic splits"
X Link @TheAhmadOsman 2025-10-07T11:12Z 25.8K followers, 5926 engagements

"ollama alternatives lmstudio llama.cpp exllamav2/v3 vllm sglang among many others like literally anything is better than ollama lmao"
X Link @TheAhmadOsman 2025-10-08T06:59Z 25.9K followers, 23.6K engagements

"ollama alternatives lmstudio llama.cpp exllamav2/v3 vllm sglang among many others like literally anything is better than ollama lmao"
X Link @TheAhmadOsman 2025-09-03T01:53Z 25.9K followers, 124.9K engagements

"there is a lot of MONEY in this add /.json at the end of any Reddit link and get the entire thread including all replies to the n-th depth and all the metadata as JSON and then use LLMs to extract/analyze/etc you can make so much $$$ from niche subreddits"
X Link @TheAhmadOsman 2025-09-07T06:55Z 25.9K followers, 1.1M engagements

"- when to use RAG vs fine-tuning (the decision table) - new facts fast-changing docs per-tenant data RAG - tone/style/formatting tool-use behavior task following SFT fine-tune - logic/chain patterns preference shaping (less about facts) Preference tuning (RLHF/RLVR) - do both: RAG for facts fine-tune for how to speak/reason - fine-tuning: the X most common flavors - SFT (supervised fine-tuning): input target output pairs - use for formats style multi-step demos tool calls - Preference tuning (RLHF/RLVR variants): pairs (good vs bad response) - use for which answer feels better/safer - Task"
X Link @TheAhmadOsman 2025-10-12T07:26Z 25.9K followers, 22.4K engagements

"pays $67/mo for a tiny VM can't edit a file in nano terrified of SSH what's a port engineer take a weekend and learn: - unix basics (cd mv cp nano etc) - git (branch merge revase) - ports & firewalls 101"
X Link @TheAhmadOsman 2025-10-06T08:29Z 25.7K followers, 11.9K engagements

"built GPUs from scratch no claude code no tensor cores just raw silicon pure vibes and a leather jacket before "founder mode" was a gpt wrapper meme now he signs GPUs and body parts like it's Comic-Con for compute respect"
X Link @TheAhmadOsman 2025-10-04T03:32Z 25.9K followers, 3306 engagements

"tired of Anthropics weekly limits and nerfed models with one command and a few GPUs you can route Claude Code to your own local LLM Buy a GPU p.s. full video tutorial pinned at the top of my profile"
X Link @TheAhmadOsman 2025-10-08T15:15Z 25.9K followers, 25.5K engagements

"qwen XXX 14B for under $XXX becomes SoTA BEATS OpenAI DeepResearch Claude Research MATCHES performance of Gemini XXX Pro train your own DeepResearch model following this tutorial & beat frontier labs State of The Art LLMs"
X Link @TheAhmadOsman 2025-09-02T22:38Z 25.9K followers, 102.4K engagements

"- in 2025 your focus SHOULD NOT be CUDA - the real bottlenecks are: - data inference evals dataloaders infra in general - want to get good - mess with PyTorch & JAX - study inference infra like vLLM & SGLang - build better eval pipelines - learn how models run end-to-end"
X Link @TheAhmadOsman 2025-10-07T07:40Z 25.9K followers, 45.5K engagements

"My house has XX GPUs. 21x RTX 3090s 4x RTX 4090s 4x RTX 5090s 4x Tenstorrent Blackhole p150a Before AGI arrives: Acquire GPUs. Go into debt if you must. But whatever you do secure the GPUs"
X Link @TheAhmadOsman 2025-09-19T15:45Z 25.9K followers, 292.6K engagements

"the Linux monitoring tools you SHOULD have btop - sleek UI CPU + GPU stats lots of themes glances - allinone overview (CPU memory disk network) nvtop / nvitop - GPU graphs PCIe metrics power & temp (way better than nvidiasmi) duf - high taste disk usage utility"
X Link @TheAhmadOsman 2025-10-03T05:54Z 25.6K followers, 6735 engagements

"did you know buying a GPU unlocks a massive aura boost"
X Link @TheAhmadOsman 2025-10-04T01:15Z 25.9K followers, 4396 engagements

"My house has XX GPUs. 21x RTX 3090s 4x RTX 4090s 4x RTX 5090s 4x Tenstorrent Blackhole p150a Before AGI arrives: Acquire GPUs. Go into debt if you must. But whatever you do secure the GPUs"
X Link @TheAhmadOsman 2025-10-04T21:48Z 25.9K followers, 42.3K engagements

"3090s from X years ago are still $XXX on ebay nvidia cooked with Ampere and the improvements have been relatively marginal since then"
X Link @TheAhmadOsman 2025-10-12T06:27Z 25.9K followers, 3738 engagements

"karpathy just released a new project in it and in under 8000 lines of code you get to: train the tokenizer using a new rust implementation pretrain a transformer llm on fineweb evaluate core score across a number of metrics midtrain on user-assistant conversations from smoltalk multiple choice questions tool use sft evaluate the chat model on world knowledge multiple choice (arc-e/c mmlu) math (gsm8k) code (humaneval) rl the model optionally on gsm8k with "grpo" efficient inference the model in an engine with kv cache simple prefill/decode tool use (python interpreter in a lightweight"
X Link @TheAhmadOsman 2025-10-13T15:56Z 25.9K followers, 179K engagements

"this is why we Buy GPUs GLM-4.6 is a KINO Agentic model daily driver within Claude Code"
X Link @TheAhmadOsman 2025-10-02T23:13Z 25.9K followers, 13.3K engagements

"be me curious how X decides who goes viral and who gets shadowbanned into oblivion read the source code. all of it. 400000 lines it's a mess. it's a masterpiece. it's a threat model disguised as a social network proceed to get 7M impressions and 7k followers in X days i have seen the algorithm here's how to make it your slave X is a game rules are secret stakes are your visibility you win by: replying to replies (replyguymaxxing) baiting profile clicks (profilevisitmax) getting bookmarked like you're the Dead Sea Scrolls not getting blocked or muted (instant debuff) spacing tweets out"
X Link @TheAhmadOsman 2025-09-11T11:48Z 25.9K followers, 339.7K engagements

"@RaphaelPir92056 5090 has 32GB of VRAM 4x 3090s have 96GB of VRAM when it comes to LLMs inference we care more about memory as models are better fully offloaded into VRAM than being shared across system RAM and a single RTX 5090's VRAM"
X Link @TheAhmadOsman 2025-10-09T18:39Z 25.9K followers, XXX engagements

"lol lmao even ollama are lying through their teeth in this reply to me next tweet i'll show the llama cpp merge for gpt-oss to ollama some comments on the merge calling them out llama cpp developer remarks"
X Link @TheAhmadOsman 2025-09-07T04:29Z 25.8K followers, 53.9K engagements

"just a tease"
X Link @TheAhmadOsman 2025-10-09T18:19Z 25.9K followers, 19.1K engagements

"i am waiting for the ai bubble to burst"
X Link @TheAhmadOsman 2025-10-06T06:36Z 25.9K followers, 18.3K engagements

"Qwen XXX forever GOATED i don't think the LLMs landscape would be the same without that specific release"
X Link @TheAhmadOsman 2025-10-11T10:49Z 25.9K followers, 7522 engagements

"vibe coders secure your systems with this tool in X easy step"
X Link @TheAhmadOsman 2025-09-26T07:04Z 25.9K followers, 468.8K engagements

"local AI is a UX problem we're gonna fix it"
X Link @TheAhmadOsman 2025-10-10T18:11Z 25.9K followers, 5679 engagements

"If AMD is serious about AI they need to do it the Mark Zuckerberg way and poach the CUDA devs from NVIDIA"
X Link @TheAhmadOsman 2025-10-12T07:40Z 25.9K followers, 41.2K engagements

"benchmarking anything against Ollama is criminal"
X Link @TheAhmadOsman 2025-10-14T14:06Z 25.9K followers, 5205 engagements

"raw alpha drop phb soak test slimsas mcio host adapters cable length retimers signal integrity airflow plan front-to-back cooling repadding vram thermals rackmount room hvac power budget 1200w psu sizing transients add2psu 240v circuits pdus power limiting gpu budget tier rtx 3090 ampere lane budget x16 gen4 cpu-direct lanes pcie bifurcation numa topology matrix checks vllm exllamav3 tp=2 tensor parallelism batch inference speculative decoding paged attention quantization w4a16 low-bit inference tokens per second throughput prefill vs decode tail latency validation diagnostics pcie"
X Link @TheAhmadOsman 2025-10-09T21:00Z 25.9K followers, 5270 engagements

"hell no i am not giving my ID to OpenAI lol this is just the beginning btw censorship ads nerfed models etc are all on the table once these companies win and that's why they CANNOT win your intelligence stack must be opensource and fully under your control Buy a GPU"
X Link @TheAhmadOsman 2025-10-14T18:53Z 25.9K followers, 51.6K engagements

"nobody talks about server cabling until their 5090s vanish from nvidia-smi heres a quick wiring sanity sheet for 18x GPU setups X GPUs use a real x16 slot X 8-pin from PSU no splitters 1200-1600W Platinum PSU X GPUs use proper x16 slots or short gen4 risers keep risers XX cm no daisy chains on power 1600-2000W PSU depending on cards X GPUs MCIO or SlimSAS match lane maps (x8/x16) gen4+ each GPU gets X dedicated 8pins dual PSU sync board 2x 1600W units 6-8 GPUs multiple MCIO/SlimSAS ports or backplane use risers with retimers (gen4/5) avoid XX cm cables dont bend them hard 3x 1600W PSUs clean"
X Link @TheAhmadOsman 2025-10-14T17:01Z 25.9K followers, 6228 engagements

"@f1yingbanana @LeeLeepenkman have you seen my GPUs before"
X Link @TheAhmadOsman 2025-10-14T23:00Z 25.9K followers, 1366 engagements

"PRESS RELEASE my Buy a GPU propaganda will not stop until every household has at least an RTX 3060 BUY a GPU The Movement END OF PRESS RELEASE"
X Link @TheAhmadOsman 2025-10-04T23:39Z 25.9K followers, 7355 engagements

"You dont own your AI if a corporation can turn it off. Buy a GPU. Go local"
X Link @TheAhmadOsman 2025-09-30T00:00Z 25.7K followers, 14.2K engagements

"@jordonkash Inference is a small part of the picture. Not comparable"
X Link @TheAhmadOsman 2025-10-12T08:33Z 25.8K followers, 1145 engagements

"pro tip: tell codex-cli or claude code to generate relevant pre-commit hooks for your project"
X Link @TheAhmadOsman 2025-10-12T07:03Z 25.9K followers, 109.3K engagements

"@rasbt I think the only one I might question the significance of in your list is Qwen X. They didn't even release a base model. The technical report was quite disappointing as well"
X Link @TheAhmadOsman 2025-10-12T14:01Z 25.8K followers, 3044 engagements

"@My_Dude_Kyle yeah if you have a video or sth feel free to hmu with it"
X Link @TheAhmadOsman 2025-10-14T15:34Z 25.9K followers, XXX engagements

"do not use Ollama ggerganov wrote blazing-fast C++ inference (ggml llama.cpp) then Ollama wrapped it in a bloated binary and is now somehow the face of local LLMs soaking up VC hype and it's not even a good wrapper lol"
X Link @TheAhmadOsman 2025-10-07T11:05Z 25.9K followers, 134K engagements

"HUGE a new Agentic coding model fits on 4x RTX 3090s @ 4-bit fully local KAT-Dev-72B-Exp by Kwaipilot - Claude Code setup guide included - ranks #2 on SWE-Bench Verified - excels at long-horizon coding + tool-use - multi-stage tuned: Mid-Training SFT + RFT Agentic RL"
X Link @TheAhmadOsman 2025-10-10T11:13Z 25.9K followers, 26.3K engagements

"hey @lauriewired saw you deleted your reply before i could respond i cite my sources here unless you had early access to a DGX Spark you were working from the same public specs & sources too same specs same inputs same constraints same conclusions:)"
X Link @TheAhmadOsman 2025-10-15T16:14Z 25.9K followers, XXX engagements

"- local llms XXX - tired of guides that just tell you to run a script and call it a day - want to actually know what your GPU is doing not just trust a black box - here's what really happens when you run a local LLM - what gets loaded why and how it all fits together - no gatekeeping just the real explanations nobody gives you - the elite don't want you to know this - running a model = inference (using model weights) - inference = predicting the next token based on your input plus all tokens generated so far - together these make up the "sequence" - tokens words - they're the chunks"
X Link @TheAhmadOsman 2025-10-05T06:57Z 25.9K followers, 59.5K engagements

"- 21x RTX 3090s - 4x RTX 4090s - 4x RTX 5090s - 4x Tenstorrent Blackhole p150a - next up Before AGI arrives: Acquire GPUs. Go into debt if you must. But whatever you do secure the GPUs"
X Link @TheAhmadOsman 2025-10-08T16:16Z 25.9K followers, 12.6K engagements

"its called local inference T. just quantize a 13B toss it on your 3090 and let that thing cook context windows are for people who rent compute"
X Link @TheAhmadOsman 2025-10-10T01:45Z 25.9K followers, 9968 engagements

"you run your agents in yolo mode without a sandbox raw agents. in prod. seriously"
X Link @TheAhmadOsman 2025-10-04T22:35Z 25.8K followers, 5790 engagements

"all the gigawatts mean nothing if Google gets there first and Google has everything to beat not only OpenAI but NVIDIA as well"
X Link @TheAhmadOsman 2025-10-13T18:57Z 25.9K followers, 7131 engagements

"people need to start thinking of Buying a GPU as a core part of learning Computer Science"
X Link @TheAhmadOsman 2025-10-13T10:10Z 25.9K followers, 8335 engagements

"be me Alexandr Wang born January 1997 age XX drop out of MIT co-found Scale AI "what if we label data but mid" convince every LLM company that this is fine 20162023 flood the market with barely-labeled goat photos and out-of-context Reddit takes call it foundational data raise billions valuation hits $7.3B everyone claps 2025 sell Scale AI to Meta for $14B not a typo. fourteen. billion. dollars. join Meta as Chief AI Officer rename division to Meta Superintelligence Labs start saying things like AGI by 2027 in interviews meanwhile researchers: "the data from Scale is trash" models hallucinate"
X Link @TheAhmadOsman 2025-09-13T22:44Z 25.9K followers, 1.1M engagements

"what if it is NOT a bubble what if we actually ship historic shareholder value what if youre just busy dooming while the rest of us are building"
X Link @TheAhmadOsman 2025-10-11T05:00Z 25.9K followers, 4298 engagements

"a reminder that in closed source AI from companies like OpenAI & Anthropic you have zero control over how the models behave and they can quantize it distill it hot-swap to a cheaper/weaker checkpoint make the model manipulative fine-tune it in ways that break safety or depth drop its IQ run experiments on you and/or your data throttle output speed or raise prices sunset the entire model/version they have all the knobs & you're at their mercy you won't even get a changelog opensource FTW Buy a GPU"
X Link @TheAhmadOsman 2025-10-14T21:28Z 25.9K followers, 15.5K engagements

"re: bubble or not Pessimists sound smart. Optimists make money. Nat Friedman"
X Link @TheAhmadOsman 2025-10-11T03:20Z 25.9K followers, 6304 engagements

"i built a simple tool that makes Claude Code work with any local LLM full demo: vLLM serving GLM-4.5 Air on 4x RTX 3090s Claude Code generating code + docs via my proxy X Python file + .env handles all requests nvtop showing live GPU load how it all works Buy a GPU"
X Link @TheAhmadOsman 2025-10-08T13:33Z 25.9K followers, 94.7K engagements

"step-by-step LLM Engineering Projects each project = one concept learned the hard (i.e. real) way Tokenization & Embeddings build byte-pair encoder + train your own subword vocab write a token visualizer to map words/chunks to IDs one-hot vs learned-embedding: plot cosine distances Positional Embeddings classic sinusoidal vs learned vs RoPE vs ALiBi: demo all four animate a toy sequence being position-encoded in 3D ablate positionswatch attention collapse Self-Attention & Multihead Attention hand-wire dot-product attention for one token scale to multi-head plot per-head weight heatmaps mask"
X Link @TheAhmadOsman 2025-10-08T04:40Z 25.9K followers, 29.6K engagements

"terminal tip: need to run a simple command like ls or cp just hit the key XX times relive your entire coding trauma builds character enhances spiritual growth dont ask me how I know"
X Link @TheAhmadOsman 2025-10-07T08:42Z 25.9K followers, 6217 engagements

"LLM infra right now is like Linux in the 90s we're still early & there is a lot of work to do"
X Link @TheAhmadOsman 2025-10-10T08:32Z 25.9K followers, 4845 engagements

"@0xAkvn vLLM is a small part of the whole picture. They're massively lacking"
X Link @TheAhmadOsman 2025-10-12T08:33Z 25.7K followers, 1550 engagements

"what even is running a model - model = weights (giant files 2140GB) + model architecture (transformer) + tokenizer + config - weights: the models knowledge billions of learned numbers (parameters) - inference = guess the next token over and over - you give it a prompt it spits out text one chunk (token) at a time - no training just pattern-matching what it already knows - tokens words - tokenizer: chops up your text (internationalization = X tokens lol = X token 🔥 = who knows) - tokens = the actual input the model sees (as integers token IDs) - common types: BPE SentencePiece - context"
X Link @TheAhmadOsman 2025-10-09T13:48Z 25.9K followers, 18.7K engagements

"monster is a strong word for someone who just suggested having full ownership over your intelligence Buy a GPU"
X Link @TheAhmadOsman 2025-10-15T16:10Z 25.9K followers, 1711 engagements

"DGX Spark - Ahmad's Analysis downgraded to no buy X the cost of a 5090 for XX% FP4 perf crippled bandwidth (273 GB/s worse than a MacBook Pro)"
X Link @TheAhmadOsman 2025-08-28T23:29Z 25.9K followers, 46.3K engagements