Loading video...

Video Failed to Load

Go Home

NVIDIA just made AI detect objects 10x faster by deleting one step. It's called LocateAnything, and it removes the biggest bottleneck no one else was fixing in vision-language models. Normally a model builds each bounding box one coordinate token at a time. 100 objects means thousands of tokens before...

dr_cintas's profile picture

Alvaro Cintas

131,752 subscribers

201,692 views • 1 month ago •via X (Twitter)

EducationNews & PoliticsScience & Technology
LIVE

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

China open-sourced a peanut-sized OCR that parses entire 100-page PDFs in one shot.. It's called Unlimited-OCR. Only 3B params. Runs locally. Every other OCR tool chops your doc into pages and loses the thread. this one reads the whole thing in a single pass. → One-shot "long-horizon" parsing (32K context window) → Multilingual, out of the box → 93% on the standard parsing benchmark (+6 over baseline) → <0.11 error rate past 40 pages → Runs 100% locally on your own hardware → Works with Transformers, vLLM, SGLang, Docker, Ollama, llama.cpp Traditional cloud OCR (Textract, Google Vision, Azure Doc Intelligence) costs $1.50–$15 per 1,000 pages. This runs on your machine. For free. Forever. Baidu built it explicitly to push DeepSeek-OCR one step further. Already at 1.9M downloads on Hugging Face and most people have no idea it exists yet. 100% open source.
0:13

China open-sourced a peanut-sized OCR that parses entire 100-page PDFs in one shot.. It's called Unlimited-OCR. Only 3B params. Runs locally. Every other OCR tool chops your doc into pages and loses the thread. this one reads the whole thing in a single pass. → One-shot "long-horizon" parsing (32K context window) → Multilingual, out of the box → 93% on the standard parsing benchmark (+6 over baseline) → <0.11 error rate past 40 pages → Runs 100% locally on your own hardware → Works with Transformers, vLLM, SGLang, Docker, Ollama, llama.cpp Traditional cloud OCR (Textract, Google Vision, Azure Doc Intelligence) costs $1.50–$15 per 1,000 pages. This runs on your machine. For free. Forever. Baidu built it explicitly to push DeepSeek-OCR one step further. Already at 1.9M downloads on Hugging Face and most people have no idea it exists yet. 100% open source.

Superman

1,099,933 views • 1 month ago

Unlimited-OCR is a 3B parameter model that parses entire 100-page PDFs in one shot — no page chunking, no lost context. +⁠32K context window, reads the whole document in a single pass +⁠93% on standard parsing benchmarks, +6 over baseline +⁠Under 0.11 error rate past page 40 — where every other OCR tool falls apart ⁠+Multilingual out of the box, runs locally via Transformers, Ollama, llama.cpp, Docker +⁠Textract, Google Vision, and Azure Doc Intelligence charge $1.50–$15 per 1,000 pages 1.9M downloads on HuggingFace and most people have never heard of it. This runs on your machine. For free. Forever. Model link:
0:13

Unlimited-OCR is a 3B parameter model that parses entire 100-page PDFs in one shot — no page chunking, no lost context. +⁠32K context window, reads the whole document in a single pass +⁠93% on standard parsing benchmarks, +6 over baseline +⁠Under 0.11 error rate past page 40 — where every other OCR tool falls apart ⁠+Multilingual out of the box, runs locally via Transformers, Ollama, llama.cpp, Docker +⁠Textract, Google Vision, and Azure Doc Intelligence charge $1.50–$15 per 1,000 pages 1.9M downloads on HuggingFace and most people have never heard of it. This runs on your machine. For free. Forever. Model link:

0xMarioNawfal

84,035 views • 1 month ago

A peanut-sized Chinese model just dethroned Gemini at reading documents. GLM-OCR is a 0.9B parameter vision-language model. It scores 94.62 on OmniDocBench V1.5, ranking #1 overall. For context, it outperforms models 100x its size. 100% open-source. It works in two stages. 1. A layout engine detects every region in a document. 2. Each region gets read in parallel. The model predicts multiple tokens per step instead of one. That's what makes it so fast at small size. It handles things most OCR tools struggle with: > Complex tables and nested layouts > Handwritten text and stamps > Math formulas and code blocks > Mixed image-and-text documents You can run it locally through Ollama. It fits on edge devices with limited compute. Every expensive OCR API just got a free competitor.
0:15

A peanut-sized Chinese model just dethroned Gemini at reading documents. GLM-OCR is a 0.9B parameter vision-language model. It scores 94.62 on OmniDocBench V1.5, ranking #1 overall. For context, it outperforms models 100x its size. 100% open-source. It works in two stages. 1. A layout engine detects every region in a document. 2. Each region gets read in parallel. The model predicts multiple tokens per step instead of one. That's what makes it so fast at small size. It handles things most OCR tools struggle with: > Complex tables and nested layouts > Handwritten text and stamps > Math formulas and code blocks > Mixed image-and-text documents You can run it locally through Ollama. It fits on edge devices with limited compute. Every expensive OCR API just got a free competitor.

AlphaSignal

92,071 views • 4 months ago

A peanut-sized Chinese model just dethroned Gemini at reading documents. GLM-OCR is a 0.9B parameter vision-language model. It scores 94.62 on OmniDocBench V1.5, ranking #1 overall. For context, it outperforms models 100x its size. 100% open-source. It works in two stages. 1. A layout engine detects every region in a document. 2. Each region gets read in parallel. The model predicts multiple tokens per step instead of one. That's what makes it so fast at small size. It handles things most OCR tools struggle with: > Complex tables and nested layouts > Handwritten text and stamps > Math formulas and code blocks > Mixed image-and-text documents You can run it locally through Ollama. It fits on edge devices with limited compute. Every expensive OCR API just got a free competitor.
0:15

A peanut-sized Chinese model just dethroned Gemini at reading documents. GLM-OCR is a 0.9B parameter vision-language model. It scores 94.62 on OmniDocBench V1.5, ranking #1 overall. For context, it outperforms models 100x its size. 100% open-source. It works in two stages. 1. A layout engine detects every region in a document. 2. Each region gets read in parallel. The model predicts multiple tokens per step instead of one. That's what makes it so fast at small size. It handles things most OCR tools struggle with: > Complex tables and nested layouts > Handwritten text and stamps > Math formulas and code blocks > Mixed image-and-text documents You can run it locally through Ollama. It fits on edge devices with limited compute. Every expensive OCR API just got a free competitor.

Jafar Najafov

13,630 views • 4 months ago

Fine-tune DeepSeek-OCR on your own language! (100% local) DeepSeek-OCR is a 3B-parameter vision model that achieves 97% precision while using 10× fewer vision tokens than text-based LLMs. It handles tables, papers, and handwriting without killing your GPU or budget. Why it matters: Most vision models treat documents as massive sequences of tokens, making long-context processing expensive and slow. DeepSeek-OCR uses context optical compression to convert 2D layouts into vision tokens, enabling efficient processing of complex documents. The best part? You can easily fine-tune it for your specific use case on a single GPU. I used Unsloth to run this experiment on Persian text and saw an 88.26% improvement in character error rate. ↳ Base model: 149% character error rate (CER) ↳ Fine-tuned model: 60% CER (57% more accurate) ↳ Training time: 60 steps on a single GPU Persian was just the test case. You can swap in your own dataset for any language, document type, or specific domain you're working with. I've shared the complete guide in the next tweet - all the code, notebooks, and environment setup ready to run with a single click. Everything is 100% open-source!
0:14

Fine-tune DeepSeek-OCR on your own language! (100% local) DeepSeek-OCR is a 3B-parameter vision model that achieves 97% precision while using 10× fewer vision tokens than text-based LLMs. It handles tables, papers, and handwriting without killing your GPU or budget. Why it matters: Most vision models treat documents as massive sequences of tokens, making long-context processing expensive and slow. DeepSeek-OCR uses context optical compression to convert 2D layouts into vision tokens, enabling efficient processing of complex documents. The best part? You can easily fine-tune it for your specific use case on a single GPU. I used Unsloth to run this experiment on Persian text and saw an 88.26% improvement in character error rate. ↳ Base model: 149% character error rate (CER) ↳ Fine-tuned model: 60% CER (57% more accurate) ↳ Training time: 60 steps on a single GPU Persian was just the test case. You can swap in your own dataset for any language, document type, or specific domain you're working with. I've shared the complete guide in the next tweet - all the code, notebooks, and environment setup ready to run with a single click. Everything is 100% open-source!

Akshay 🚀

126,122 views • 9 months ago

Qwen3.6 35B A3B can't fill out a paper form on its own. But give it NVIDIA's LocateAnything-3B — the #1 trending model on HuggingFace — as its eyes, and the two small models get it done together. (The test: place each element at the right pixel position on a blank form image, not type into a field.) Setup: > Qwen is the brain (main model), LocateAnything is the eyes (helper model acting as a tool). > I gave Qwen a new tool: ask "where's the email field?" and LocateAnything returns the exact x, y, width, height. > The blue boxes on the screen are its detections. Look how tight they are — it nails every field. Result: > Qwen3.6 35B A3B + LocateAnything-3B: form completed, all info correct. > Name, DOB, ID, gender, marital status, nationality, email, phone, address, postal code: all landed in the right field areas. > Character-box alignment still a touch loose, but every value is where it belongs. > 9m10s, 224.5k input, 24.3k output, 21 turns. Why it matters: > Qwen alone can't finish this test. Bolt on a 3B model that does exactly one thing > locate > and suddenly it can. > A combination of small models can do the work of a single large one.
0:14

Qwen3.6 35B A3B can't fill out a paper form on its own. But give it NVIDIA's LocateAnything-3B — the #1 trending model on HuggingFace — as its eyes, and the two small models get it done together. (The test: place each element at the right pixel position on a blank form image, not type into a field.) Setup: > Qwen is the brain (main model), LocateAnything is the eyes (helper model acting as a tool). > I gave Qwen a new tool: ask "where's the email field?" and LocateAnything returns the exact x, y, width, height. > The blue boxes on the screen are its detections. Look how tight they are — it nails every field. Result: > Qwen3.6 35B A3B + LocateAnything-3B: form completed, all info correct. > Name, DOB, ID, gender, marital status, nationality, email, phone, address, postal code: all landed in the right field areas. > Character-box alignment still a touch loose, but every value is where it belongs. > 9m10s, 224.5k input, 24.3k output, 21 turns. Why it matters: > Qwen alone can't finish this test. Bolt on a 3B model that does exactly one thing > locate > and suddenly it can. > A combination of small models can do the work of a single large one.

stevibe

150,373 views • 2 months ago

AN ANTHROPIC SYSTEM TURNS ONE PROMPT INTO A COMPLETE GRAPH MODEL BUILT ENTIRELY ON ITS OWN One prompt is enough to spin up the full architecture, no schema written by hand, no roles assigned in advance. Parallel workers process every branch of the graph at once, each one handling its own slice independently. As each worker finishes, its output compacts straight into a single unified project instead of scattered fragments. Nothing needs stitching together manually, the compaction step folds every worker's result into one file automatically. The whole build runs end to end without a single manual step between the prompt and the finished project. See the full build and compact pipeline below👇
0:15

AN ANTHROPIC SYSTEM TURNS ONE PROMPT INTO A COMPLETE GRAPH MODEL BUILT ENTIRELY ON ITS OWN One prompt is enough to spin up the full architecture, no schema written by hand, no roles assigned in advance. Parallel workers process every branch of the graph at once, each one handling its own slice independently. As each worker finishes, its output compacts straight into a single unified project instead of scattered fragments. Nothing needs stitching together manually, the compaction step folds every worker's result into one file automatically. The whole build runs end to end without a single manual step between the prompt and the finished project. See the full build and compact pipeline below👇

wast3

18,412 views • 22 days ago

JENSEN HUANG UNVEILED A BOARD THAT RUNS 1 TRILLION PARAMETER AI MODELS. THE $249 NVIDIA BOX UNDER YOUR DESK KILLS A $200/MONTH AI BILL FOR $5 IN ELECTRICITY jensen held it up on stage with one hand and called it the architecture that runs the future of ai. that same technology now ships in a $249 box smaller than your wallet the jetson orin nano super pulls 7-25 watts and does 67 trillion ai operations per second. llama 3, mistral and deepseek run locally with no api fees and no data leaving your machine most developers pay $2,400 a year across chatgpt, openai api, claude pro and cursor. the jetson costs $314 in year one and $60 a year after. 2 year savings hit $4,431 install ollama with one command, change one line of code to point at localhost, and every tool built for openai works identically. zero rewrites, zero rate limits cloud subscriptions keep getting more expensive and rate limits keep getting tighter. the people who own the box in 2026 are going to look very far ahead in 2028 bookmark this and read the article below
0:15

JENSEN HUANG UNVEILED A BOARD THAT RUNS 1 TRILLION PARAMETER AI MODELS. THE $249 NVIDIA BOX UNDER YOUR DESK KILLS A $200/MONTH AI BILL FOR $5 IN ELECTRICITY jensen held it up on stage with one hand and called it the architecture that runs the future of ai. that same technology now ships in a $249 box smaller than your wallet the jetson orin nano super pulls 7-25 watts and does 67 trillion ai operations per second. llama 3, mistral and deepseek run locally with no api fees and no data leaving your machine most developers pay $2,400 a year across chatgpt, openai api, claude pro and cursor. the jetson costs $314 in year one and $60 a year after. 2 year savings hit $4,431 install ollama with one command, change one line of code to point at localhost, and every tool built for openai works identically. zero rewrites, zero rate limits cloud subscriptions keep getting more expensive and rate limits keep getting tighter. the people who own the box in 2026 are going to look very far ahead in 2028 bookmark this and read the article below

starmex

54,448 views • 2 months ago

Baidu just open-sourced an OCR model that reads entire 40-page documents in one shot. It's called Unlimited-OCR. 3 billion parameters but only 500 million active during inference. Runs 100% locally on your machine. Why this matters: traditional OCR tools chop documents page by page. Tables that span two pages break. Reading order gets lost. Cross-page context disappears. Unlimited-OCR processes the whole document at once. 32K context window. Text, formulas, tables, reading order all preserved across pages. Output comes out as clean structured Markdown. → 93% accuracy on the standard benchmark. +6 points over the baseline. → Error rate stays below 0.11 even past 40 pages. → Multilingual out of the box. → 2.12 million downloads on Hugging Face last month. 14,600 GitHub stars. For context: Amazon Textract, Google Cloud Vision, and Azure Document Intelligence all charge per page. This runs locally for free.
0:13

Baidu just open-sourced an OCR model that reads entire 40-page documents in one shot. It's called Unlimited-OCR. 3 billion parameters but only 500 million active during inference. Runs 100% locally on your machine. Why this matters: traditional OCR tools chop documents page by page. Tables that span two pages break. Reading order gets lost. Cross-page context disappears. Unlimited-OCR processes the whole document at once. 32K context window. Text, formulas, tables, reading order all preserved across pages. Output comes out as clean structured Markdown. → 93% accuracy on the standard benchmark. +6 points over the baseline. → Error rate stays below 0.11 even past 40 pages. → Multilingual out of the box. → 2.12 million downloads on Hugging Face last month. 14,600 GitHub stars. For context: Amazon Textract, Google Cloud Vision, and Azure Document Intelligence all charge per page. This runs locally for free.

Vaibhav Sisinty

414,555 views • 1 month ago

sorry, they just did WHAT someone gave a machine one disease name, the leading cause of blindness in the developed world with 1.5 million americans already in its path, and it came back pointing at a drug that has sat in pharmacies for years under a different label: 551 papers read in 30 minutes against the 294 hours a human would have needed, and the loop that did it is public on GitHub most agent setups answer one question at a time, so the ceiling on the work is the quality of the question you happened to think of this one was handed a single question and wrote the second one itself. turns out that follow-up is where the real find was: a target called ABCA1, upregulated threefold, in an experiment no human ordered i read the whole paper looking for the trick, and the trick is structural. that is the second question, and it is the gap between an assistant and a factory: - hand the loop a field rather than a task: it was given a disease, and choosing the mechanism was part of its job - make it rank before it spends: 151 papers in, ten candidate mechanisms out, scored against each other before anything touched a bench - split reading from judging, so the agent that forms the theory is a different agent from the one grading it - close every cycle on physical reality: the verdict was an experiment, and another model's opinion was never allowed to stand in for one - feed each result back as the next question rather than a log line, which is the step almost nobody builds - search what already passed inspection first: the winner was an approved compound with a safety file already on record - write down what the round learned before opening the next one, so round two starts where round one stopped my read, and i think it is the uncomfortable one: reading was the entire bottleneck in that field, and everybody spent the decade optimising the writing. people ran every physical experiment here, the analysis agent needs a domain expert writing its prompts, and the authors decline to call this the leap it resembles. the thinking got replaced, and the hands did not so the question i cannot answer for my own setup: which step of your loop still stops dead until you sit down and type something bookmark this one. the four parts that turn one model into a line that runs like this, the queue, the rooms, the write permissions and the gate, are built file by file in the piece below ↓
0:15

sorry, they just did WHAT someone gave a machine one disease name, the leading cause of blindness in the developed world with 1.5 million americans already in its path, and it came back pointing at a drug that has sat in pharmacies for years under a different label: 551 papers read in 30 minutes against the 294 hours a human would have needed, and the loop that did it is public on GitHub most agent setups answer one question at a time, so the ceiling on the work is the quality of the question you happened to think of this one was handed a single question and wrote the second one itself. turns out that follow-up is where the real find was: a target called ABCA1, upregulated threefold, in an experiment no human ordered i read the whole paper looking for the trick, and the trick is structural. that is the second question, and it is the gap between an assistant and a factory: - hand the loop a field rather than a task: it was given a disease, and choosing the mechanism was part of its job - make it rank before it spends: 151 papers in, ten candidate mechanisms out, scored against each other before anything touched a bench - split reading from judging, so the agent that forms the theory is a different agent from the one grading it - close every cycle on physical reality: the verdict was an experiment, and another model's opinion was never allowed to stand in for one - feed each result back as the next question rather than a log line, which is the step almost nobody builds - search what already passed inspection first: the winner was an approved compound with a safety file already on record - write down what the round learned before opening the next one, so round two starts where round one stopped my read, and i think it is the uncomfortable one: reading was the entire bottleneck in that field, and everybody spent the decade optimising the writing. people ran every physical experiment here, the analysis agent needs a domain expert writing its prompts, and the authors decline to call this the leap it resembles. the thinking got replaced, and the hands did not so the question i cannot answer for my own setup: which step of your loop still stops dead until you sit down and type something bookmark this one. the four parts that turn one model into a line that runs like this, the queue, the rooms, the write permissions and the gate, are built file by file in the piece below ↓

Argona

32,475 views • 17 days ago

THIS SHELF OF MAC MINIS REPLACES $4,080 A YEAR IN AI SUBSCRIPTIONS 00:02 the camera pans across a shelf of stacked Mac minis and the trick is obvious: that silent little farm runs the models you rent every month most people pay 7 companies for AI and use 3 of the tools. they forget the rest on the credit card and call it a stack the Mac mini M4 ends that. one shared memory pool means a $599 box runs 7B and 8B models faster than Windows machines that cost twice as much ollama pull, one command. open webui in one docker line. point Claude Code at localhost and it just works it draws 10 to 30 watts, sits silent next to a router, and runs 24/7 for $3 a month in power it pays back a $20 ChatGPT Plus sub in 3 months, then saves you $4,000 a year while the frontier still rents you compute every month you wait is another $340 gone for compute that fits on a shelf
0:10

THIS SHELF OF MAC MINIS REPLACES $4,080 A YEAR IN AI SUBSCRIPTIONS 00:02 the camera pans across a shelf of stacked Mac minis and the trick is obvious: that silent little farm runs the models you rent every month most people pay 7 companies for AI and use 3 of the tools. they forget the rest on the credit card and call it a stack the Mac mini M4 ends that. one shared memory pool means a $599 box runs 7B and 8B models faster than Windows machines that cost twice as much ollama pull, one command. open webui in one docker line. point Claude Code at localhost and it just works it draws 10 to 30 watts, sits silent next to a router, and runs 24/7 for $3 a month in power it pays back a $20 ChatGPT Plus sub in 3 months, then saves you $4,000 a year while the frontier still rents you compute every month you wait is another $340 gone for compute that fits on a shelf

Fokki

12,933 views • 1 month ago

LTX 2.5 JUST WENT FASTER THAN REAL TIME LTX (LTX.io) is already one of the world's most widely used open video models. We just made its latest release, LTX 2.5, fast enough for rapid iteration and affordable enough for production at scale. It’s live on TensorScaleAI now, generating a 5-sec 1080p video in just 3.12s at $0.04 per second. That's faster than real time, at a fraction of the cost! For creative teams, that means less waiting time, more ideas explored, and lower production costs. For robotics and physical AI teams, that means more data generated at scale, more scenarios tested, and shorter development and deployment cycles. Request early access today and we’ll get you onboarded! (link:
0:12

LTX 2.5 JUST WENT FASTER THAN REAL TIME LTX (LTX.io) is already one of the world's most widely used open video models. We just made its latest release, LTX 2.5, fast enough for rapid iteration and affordable enough for production at scale. It’s live on TensorScaleAI now, generating a 5-sec 1080p video in just 3.12s at $0.04 per second. That's faster than real time, at a fraction of the cost! For creative teams, that means less waiting time, more ideas explored, and lower production costs. For robotics and physical AI teams, that means more data generated at scale, more scenarios tested, and shorter development and deployment cycles. Request early access today and we’ll get you onboarded! (link:

Yao Zhang

15,425 views • 6 days ago

AI has had exactly two scaling axes that worked so far, and the second one is starting to look finite too the first one was pretraining: with scaling parameters and data, we got world knowledge (i.e. ChatGPT had read enough to know things), but it started saturating a while ago the second one was RL, and people had been doing RL the whole time before that: RLHF is RL but it never scaled far because it was trying to control the exact output, which tokens come out, how the text reads, but you can only push that so far before you’re just polishing RLVR dropped that constraint: giving the model a task, then checking whether the final answer is right, and ignoring everything in between -- so the model does whatever it wants in the middle and only the endpoint gets graded, and that’s much closer to actual RL and it’s what bought us planning and reasoning (arguably, tool use sits around 2.5 on this list -- while useful, it's not a different kind of thing) so one axis gave knowledge, the other gave reasoning, and both of them are one model working alone the next axis is how many models you can get working on the same problem, which is a different kind of axis than the previous two we know that multi-agent RL has always been the harder problem: I spent years in that literature and the gap between single-agent and multi-agent is definitely not incremental -- it’s a whole different class of difficulty! which is also why the derivatives are steep at the start, nobody has picked the easy wins yet... and the thing that gates this multi-agent coordination is communication: models can only coordinate as well as they can exchange information, and right now they do that by writing sentences to each other imagine what could we possibly achieve if we properly open that third axis development by letting models to exchange information in their native "language" without loosing any computational data that they produce during inference
0:14

AI has had exactly two scaling axes that worked so far, and the second one is starting to look finite too the first one was pretraining: with scaling parameters and data, we got world knowledge (i.e. ChatGPT had read enough to know things), but it started saturating a while ago the second one was RL, and people had been doing RL the whole time before that: RLHF is RL but it never scaled far because it was trying to control the exact output, which tokens come out, how the text reads, but you can only push that so far before you’re just polishing RLVR dropped that constraint: giving the model a task, then checking whether the final answer is right, and ignoring everything in between -- so the model does whatever it wants in the middle and only the endpoint gets graded, and that’s much closer to actual RL and it’s what bought us planning and reasoning (arguably, tool use sits around 2.5 on this list -- while useful, it's not a different kind of thing) so one axis gave knowledge, the other gave reasoning, and both of them are one model working alone the next axis is how many models you can get working on the same problem, which is a different kind of axis than the previous two we know that multi-agent RL has always been the harder problem: I spent years in that literature and the gap between single-agent and multi-agent is definitely not incremental -- it’s a whole different class of difficulty! which is also why the derivatives are steep at the start, nobody has picked the easy wins yet... and the thing that gates this multi-agent coordination is communication: models can only coordinate as well as they can exchange information, and right now they do that by writing sentences to each other imagine what could we possibly achieve if we properly open that third axis development by letting models to exchange information in their native "language" without loosing any computational data that they produce during inference

Sasha Malysheva

11,393 views • 14 days ago

Big win for open-source LLMs! DeepSeek V4 Pro holds the top open-weights score on SWE-bench Verified, in the GPT-5.5 range. GLM 5.2 leads the open-weight intelligence index and sits near the closed frontier on long-horizon coding. But this leaderboard number is a weak proxy for real performance. It comes from one task set, run through one harness, served at one precision. The same weights can even score differently across providers, since many hosts quantize activations to fp8 and drift the model off its reference weights. Real performance is determined based on whether a model can read a repo, make coordinated edits across files, run the tests, and recover when one breaks. By that measure, the top open models hold up, but only inside the right harness. The teams that actually put DeepSeek V4 into production pipelines as a frontier substitute got there through the harness they built around the model, not by picking a stronger model. If you want to see this in practice, Cline (64k+ stars) has actually built that harness around open models, tuned so they run at production quality. And it's tuned so that these LLMs can run at production quality, with plan and act modes, checkpoints, and terminal feedback. ClinePass is the new access layer on top of it. It runs a curated set of those models inside Cline, narrowed to the ones tested for coding-agent use, with 2 to 5x the standard rate limits and no separate provider accounts, keys, or billing to track. The video below shows the setup, and I worked with the team to put this together. It runs alongside custom keys and local models as well, not in place of them.
0:14

Big win for open-source LLMs! DeepSeek V4 Pro holds the top open-weights score on SWE-bench Verified, in the GPT-5.5 range. GLM 5.2 leads the open-weight intelligence index and sits near the closed frontier on long-horizon coding. But this leaderboard number is a weak proxy for real performance. It comes from one task set, run through one harness, served at one precision. The same weights can even score differently across providers, since many hosts quantize activations to fp8 and drift the model off its reference weights. Real performance is determined based on whether a model can read a repo, make coordinated edits across files, run the tests, and recover when one breaks. By that measure, the top open models hold up, but only inside the right harness. The teams that actually put DeepSeek V4 into production pipelines as a frontier substitute got there through the harness they built around the model, not by picking a stronger model. If you want to see this in practice, Cline (64k+ stars) has actually built that harness around open models, tuned so they run at production quality. And it's tuned so that these LLMs can run at production quality, with plan and act modes, checkpoints, and terminal feedback. ClinePass is the new access layer on top of it. It runs a curated set of those models inside Cline, narrowed to the ones tested for coding-agent use, with 2 to 5x the standard rate limits and no separate provider accounts, keys, or billing to track. The video below shows the setup, and I worked with the team to put this together. It runs alongside custom keys and local models as well, not in place of them.

Avi Chawla

44,124 views • 1 month ago

34,547 ELEMENTS. 34 WARNINGS. 7 DUPLICATE ELEMENTS REVIT'S OWN SCHEDULE NEVER CAUGHT. One prompt, one live Revit model, one Claude session through the MCP connection. No plugins, no exported schedules to reformat by hand. Claude walked the model tree: rooms, families, sheets, every open warning, cross-checked counts against each other, and came back with a full audit. Top flag: three fire-rated doors modeled without a fire rating parameter. The rest ranked below it, from code-adjacent to cosmetic. Still the read-only side of Revit's official MCP server, the part Autodesk shipped first, on purpose. Write access is on their own roadmap, through a separate server built specifically for it. Nothing here moved a single element. The model just got properly looked at, faster than a manual QA pass usually runs.
0:19

34,547 ELEMENTS. 34 WARNINGS. 7 DUPLICATE ELEMENTS REVIT'S OWN SCHEDULE NEVER CAUGHT. One prompt, one live Revit model, one Claude session through the MCP connection. No plugins, no exported schedules to reformat by hand. Claude walked the model tree: rooms, families, sheets, every open warning, cross-checked counts against each other, and came back with a full audit. Top flag: three fire-rated doors modeled without a fire rating parameter. The rest ranked below it, from code-adjacent to cosmetic. Still the read-only side of Revit's official MCP server, the part Autodesk shipped first, on purpose. Write access is on their own roadmap, through a separate server built specifically for it. Nothing here moved a single element. The model just got properly looked at, faster than a manual QA pass usually runs.

Solvaix

28,689 views • 24 days ago

500 agents checking each other isn't 500 checks. It's 124,750. That's the swarm running above. Parallel, open-source fed, cross-verifying. Nobody prices the cross. Every pair is one comparison. The count grows on the square. You added one agent and bought 499. And all of them read the same repo. One README, 500 times. You pay per reader, not per document. Grok 4.6 bills $2 per million in, $6 out. Cheap per token. Not cheap per 124,750. Grok Build spawns eight sub-agents per prompt. 500 means an orchestrator you wrote yourself. Nobody audits their own orchestrator. xAI lost $2.47 billion last quarter. Revenue was $818 million. Anthropic pays them $1.25 billion a month for compute. Your swarm runs on rented losses. Consensus never needed everybody. One shared read. Three checkers. One arbiter. Redundancy is not verification. Don't let this rot in your bookmarks. Save it, then count the pairs in your own swarm before you scale it - because 500 agents reading one file don't verify anything, they agree 500 times.
0:14

500 agents checking each other isn't 500 checks. It's 124,750. That's the swarm running above. Parallel, open-source fed, cross-verifying. Nobody prices the cross. Every pair is one comparison. The count grows on the square. You added one agent and bought 499. And all of them read the same repo. One README, 500 times. You pay per reader, not per document. Grok 4.6 bills $2 per million in, $6 out. Cheap per token. Not cheap per 124,750. Grok Build spawns eight sub-agents per prompt. 500 means an orchestrator you wrote yourself. Nobody audits their own orchestrator. xAI lost $2.47 billion last quarter. Revenue was $818 million. Anthropic pays them $1.25 billion a month for compute. Your swarm runs on rented losses. Consensus never needed everybody. One shared read. Three checkers. One arbiter. Redundancy is not verification. Don't let this rot in your bookmarks. Save it, then count the pairs in your own swarm before you scale it - because 500 agents reading one file don't verify anything, they agree 500 times.

wast3

17,964 views • 4 days ago

Holy shit... Microsoft open sourced an inference framework that runs a 100B parameter LLM on a single CPU. It's called BitNet. And it does what was supposed to be impossible. No GPU. No cloud. No $10K hardware setup. Just your laptop running a 100-billion parameter model at human reading speed. Here's how it works: Every other LLM stores weights in 32-bit or 16-bit floats. BitNet uses 1.58 bits. Weights are ternary just -1, 0, or +1. That's it. No floats. No expensive matrix math. Pure integer operations your CPU was already built for. The result: - 100B model runs on a single CPU at 5-7 tokens/second - 2.37x to 6.17x faster than llama.cpp on x86 - 82% lower energy consumption on x86 CPUs - 1.37x to 5.07x speedup on ARM (your MacBook) - Memory drops by 16-32x vs full-precision models The wildest part: Accuracy barely moves. BitNet b1.58 2B4T their flagship model was trained on 4 trillion tokens and benchmarks competitively against full-precision models of the same size. The quantization isn't destroying quality. It's just removing the bloat. What this actually means: - Run AI completely offline. Your data never leaves your machine - Deploy LLMs on phones, IoT devices, edge hardware - No more cloud API bills for inference - AI in regions with no reliable internet The model supports ARM and x86. Works on your MacBook, your Linux box, your Windows machine. 27.4K GitHub stars. 2.2K forks. Built by Microsoft Research. 100% Open Source. MIT License.
0:13

Holy shit... Microsoft open sourced an inference framework that runs a 100B parameter LLM on a single CPU. It's called BitNet. And it does what was supposed to be impossible. No GPU. No cloud. No $10K hardware setup. Just your laptop running a 100-billion parameter model at human reading speed. Here's how it works: Every other LLM stores weights in 32-bit or 16-bit floats. BitNet uses 1.58 bits. Weights are ternary just -1, 0, or +1. That's it. No floats. No expensive matrix math. Pure integer operations your CPU was already built for. The result: - 100B model runs on a single CPU at 5-7 tokens/second - 2.37x to 6.17x faster than llama.cpp on x86 - 82% lower energy consumption on x86 CPUs - 1.37x to 5.07x speedup on ARM (your MacBook) - Memory drops by 16-32x vs full-precision models The wildest part: Accuracy barely moves. BitNet b1.58 2B4T their flagship model was trained on 4 trillion tokens and benchmarks competitively against full-precision models of the same size. The quantization isn't destroying quality. It's just removing the bloat. What this actually means: - Run AI completely offline. Your data never leaves your machine - Deploy LLMs on phones, IoT devices, edge hardware - No more cloud API bills for inference - AI in regions with no reliable internet The model supports ARM and x86. Works on your MacBook, your Linux box, your Windows machine. 27.4K GitHub stars. 2.2K forks. Built by Microsoft Research. 100% Open Source. MIT License.

Guri Singh

2,180,357 views • 5 months ago

19-year-old from china makes $9,000/month designing product sites and ships each one in an afternoon. here's his exact setup the whole thing runs on two tools that each do one job: > brief written by hand: 5 min > Moonchild builds the design system, then every screen from it: 20 min > MCP hands the design to Claude as real structure, not a screenshot: instant > Claude Code reads those exact tokens and builds the live app: 20 min > second Claude session reviews the build for drift: 10 min total: about an hour. screen five still matches screen one. no agency, no dev, no design team the trick is MCP. the design tool passes Claude the actual colors, components and layout, so it builds from the source instead of guessing from a picture. full pipeline, every prompt, in the article above.
0:12

19-year-old from china makes $9,000/month designing product sites and ships each one in an afternoon. here's his exact setup the whole thing runs on two tools that each do one job: > brief written by hand: 5 min > Moonchild builds the design system, then every screen from it: 20 min > MCP hands the design to Claude as real structure, not a screenshot: instant > Claude Code reads those exact tokens and builds the live app: 20 min > second Claude session reviews the build for drift: 10 min total: about an hour. screen five still matches screen one. no agency, no dev, no design team the trick is MCP. the design tool passes Claude the actual colors, components and layout, so it builds from the source instead of guessing from a picture. full pipeline, every prompt, in the article above.

Ridark

19,477 views • 2 months ago

A PACK OF ROBOT DOGS JUST RAN A GROUP MOBILITY TEST THROUGH MUD AND STANDING WATER. Black quadrupeds, marked G01, tactical operators in black gear crouched around one on a wet, overcast field before the run starts. Then the label hits: RECON UNIT MOBILITY TEST — GROUP RUN. Several units take off together across churned mud and puddles. They don’t break formation. No slipping, no falling behind, no unit losing its footing on the uneven ground. This isn’t a single robot proving it can walk. It’s a coordinated group run — the kind of test that matters for scouting terrain a human patrol would move through slower and louder. The honest catch: bulkier, more armored builds than consumer platforms like Unitree’s Go1 usually mean less speed and battery life traded for durability. This isn’t a sprinting demo, it’s a ruggedness one. Ten years ago “robot dog” meant a single unit trotting across a clean lab floor. This is a pack of them holding formation through mud a human squad would curse at. Would a mud field even slow one of these down?
0:16

A PACK OF ROBOT DOGS JUST RAN A GROUP MOBILITY TEST THROUGH MUD AND STANDING WATER. Black quadrupeds, marked G01, tactical operators in black gear crouched around one on a wet, overcast field before the run starts. Then the label hits: RECON UNIT MOBILITY TEST — GROUP RUN. Several units take off together across churned mud and puddles. They don’t break formation. No slipping, no falling behind, no unit losing its footing on the uneven ground. This isn’t a single robot proving it can walk. It’s a coordinated group run — the kind of test that matters for scouting terrain a human patrol would move through slower and louder. The honest catch: bulkier, more armored builds than consumer platforms like Unitree’s Go1 usually mean less speed and battery life traded for durability. This isn’t a sprinting demo, it’s a ruggedness one. Ten years ago “robot dog” meant a single unit trotting across a clean lab floor. This is a pack of them holding formation through mud a human squad would curse at. Would a mud field even slow one of these down?

DN_DEGEN

59,530 views • 20 days ago

whoever leaked this has bigger balls than sense someone gave a fleet of Claude agents shared memory so they would stop contradicting each other, then measured both the bill and the output: the version that talked most made 2.4x the api calls of the version that won, and hallucinated 34% more than doing nothing at all, 0.658 against 0.492 i ran the same question past two of my own agents afterwards and got two different answers about which file owns the config. each one was individually right and the pair was wrong, which is the whole failure in one line this is Graph Engineering, the layer that decides which agents may talk to each other at all, and it installs into the agent you already pay for: - decide which agents may share state at all, because every edge you draw is a channel a mistake can travel down - measure divergence per PAIR instead of as a fleet average, across what they believe about place, time and task history - gate on that number and stop the pair above your threshold before it reasons, rather than repairing the output afterwards - let compressed summaries replace whole states: the verified protocol landed 0.463 against 0.658 for full broadcast - cut the sync frequency until it hurts, since the winning setup used 58% fewer calls than the one that broke it - never propagate a state nobody checked, because the contamination effect came in at d=1.18, a full standard deviation of extra lying - keep the shared layer small enough to diff, which is what a written standard does and a running conversation cannot - re-run the check after every model upgrade, because this was 8 scenarios on one model family at n=30 per condition - and learn where it does not bite: on plain software tasks every condition converged under 0.2 and the whole effect vanished turns out the ranking is the uncomfortable part: verified summaries 0.463, no synchronisation at all 0.492, full broadcast 0.658. the middle option is doing nothing, and it beat the thing everyone builds first the group agreeing is what it looks like when every agent copied the same mistake, which is why a fleet that hallucinates has a replication problem and keeps getting handed a smarter model instead so the question for your own setup: if you asked two of your agents the same thing right now, would they answer the same way bookmark this one. the layer underneath it, deciding which arrows between agents exist at all, is built step by step in the piece below ↓
0:15

whoever leaked this has bigger balls than sense someone gave a fleet of Claude agents shared memory so they would stop contradicting each other, then measured both the bill and the output: the version that talked most made 2.4x the api calls of the version that won, and hallucinated 34% more than doing nothing at all, 0.658 against 0.492 i ran the same question past two of my own agents afterwards and got two different answers about which file owns the config. each one was individually right and the pair was wrong, which is the whole failure in one line this is Graph Engineering, the layer that decides which agents may talk to each other at all, and it installs into the agent you already pay for: - decide which agents may share state at all, because every edge you draw is a channel a mistake can travel down - measure divergence per PAIR instead of as a fleet average, across what they believe about place, time and task history - gate on that number and stop the pair above your threshold before it reasons, rather than repairing the output afterwards - let compressed summaries replace whole states: the verified protocol landed 0.463 against 0.658 for full broadcast - cut the sync frequency until it hurts, since the winning setup used 58% fewer calls than the one that broke it - never propagate a state nobody checked, because the contamination effect came in at d=1.18, a full standard deviation of extra lying - keep the shared layer small enough to diff, which is what a written standard does and a running conversation cannot - re-run the check after every model upgrade, because this was 8 scenarios on one model family at n=30 per condition - and learn where it does not bite: on plain software tasks every condition converged under 0.2 and the whole effect vanished turns out the ranking is the uncomfortable part: verified summaries 0.463, no synchronisation at all 0.492, full broadcast 0.658. the middle option is doing nothing, and it beat the thing everyone builds first the group agreeing is what it looks like when every agent copied the same mistake, which is why a fleet that hallucinates has a replication problem and keeps getting handed a smarter model instead so the question for your own setup: if you asked two of your agents the same thing right now, would they answer the same way bookmark this one. the layer underneath it, deciding which arrows between agents exist at all, is built step by step in the piece below ↓

Argona

722,569 views • 15 days ago