Loading video...

Video Failed to Load

Go Home

Single vs Multi-hand Attention by hand ✍️ Resize matrices yourself 👉 The most important fact about multi-head attention: it has the same parameter count as single-head attention. The difference is purely structural — same total Wqkv weights, partitioned into smaller q–k–v triples. Look at the two diagrams below. Both...

ProfTomYeh's profile picture

Tom Yeh

64,669 subscribers

35,807 views • 3 months ago •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

Self Attention vs Cross Attention by hand ✍️ Resize the matrices yourself 👉 Two attention mechanisms, side by side. Both project X into queries; both compute attention via S = Kᵀ × Q and F = V × A. The only difference is the source of K and V. Self attention uses X for everything. Q, K, and V all come from projecting X. Each X token attends to every other X token. The score matrix S is square — 128 × 128. Cross attention uses X for queries and a second sequence E for keys and values. Each X token attends to every E token instead. The score matrix S is rectangular — 64 × 128. Notice what's shared and what's not: X is the same in both — same 36 × 128 input. Q and K share the 16 dimension — that's what makes the dot product Kᵀ × Q valid in either case. V dimensions are independent: self-attention uses 12, cross-attention uses 12. The choice doesn't depend on which mechanism you're using; it depends on what output dimension your downstream layer expects.
0:18

Self Attention vs Cross Attention by hand ✍️ Resize the matrices yourself 👉 Two attention mechanisms, side by side. Both project X into queries; both compute attention via S = Kᵀ × Q and F = V × A. The only difference is the source of K and V. Self attention uses X for everything. Q, K, and V all come from projecting X. Each X token attends to every other X token. The score matrix S is square — 128 × 128. Cross attention uses X for queries and a second sequence E for keys and values. Each X token attends to every E token instead. The score matrix S is rectangular — 64 × 128. Notice what's shared and what's not: X is the same in both — same 36 × 128 input. Q and K share the 16 dimension — that's what makes the dot product Kᵀ × Q valid in either case. V dimensions are independent: self-attention uses 12, cross-attention uses 12. The choice doesn't depend on which mechanism you're using; it depends on what output dimension your downstream layer expects.

Tom Yeh

61,300 views • 4 months ago

Transformer by hand ✍️ ~ 6 steps walkthrough below Open the hood of a transformer and the parts list is overwhelming: embeddings, positional encoding, attention weighting, self-attention, cross-attention, multi-head attention, layer norm, skip connections, softmax, linear, Nx, shifted right, query, key, value, masking. Which of those actually make the car run? Two of them. Attention weighting and the feed-forward network. Everything else is an enhancement to make it run faster and longer, which is how we got from a car to a truck, and to the word "large" in large language model. So I drew and calculated those two parts entirely by hand. Goal: push five features through one transformer block, filling in every cell yourself. 1. Given Five positions of input features, arriving from the previous block. 2. Attention matrix Let us feed all five features to a query-key module (QK) and read back an attention weight matrix, A. The details of that module are a post of their own. 3. Attention weighting We multiply the input features by A to get the attention weighted features, Z. Still five positions. The effect is to combine features *across positions*, horizontally: X1 becomes X1 + X2, X2 becomes X2 + X3, and so on. 4. First layer Let us feed all five weighted features into the first layer of the FFN. Multiply by the weights and biases. This time the combining happens *across feature dimensions*, vertically, and each feature grows from 3 numbers to 4. Note that every position goes through the same weight matrix. That is what "position-wise" means. 5. ReLU We cross out the negatives. They become zeros. 6. Second layer Let us bring it back down: 4 dimensions to 3. The output feeds the next block, which has a completely separate set of parameters, and the whole thing runs again. You have just calculated a transformer block by hand. ✍️ The takeaway: the two parts are doing two different jobs, and neither one alone is enough. Attention mixes *across positions*, so a feature can see its neighbours. The FFN mixes *across feature dimensions*, so each position can think about itself. Horizontal, then vertical. Then that pattern repeats N times, each block with its own separate set of weights. That is the Nx from the list up top, and that is what makes the transformer run. 💾 Save this post! #AIbyHand #Transformers #DeepLearning
0:12

Transformer by hand ✍️ ~ 6 steps walkthrough below Open the hood of a transformer and the parts list is overwhelming: embeddings, positional encoding, attention weighting, self-attention, cross-attention, multi-head attention, layer norm, skip connections, softmax, linear, Nx, shifted right, query, key, value, masking. Which of those actually make the car run? Two of them. Attention weighting and the feed-forward network. Everything else is an enhancement to make it run faster and longer, which is how we got from a car to a truck, and to the word "large" in large language model. So I drew and calculated those two parts entirely by hand. Goal: push five features through one transformer block, filling in every cell yourself. 1. Given Five positions of input features, arriving from the previous block. 2. Attention matrix Let us feed all five features to a query-key module (QK) and read back an attention weight matrix, A. The details of that module are a post of their own. 3. Attention weighting We multiply the input features by A to get the attention weighted features, Z. Still five positions. The effect is to combine features *across positions*, horizontally: X1 becomes X1 + X2, X2 becomes X2 + X3, and so on. 4. First layer Let us feed all five weighted features into the first layer of the FFN. Multiply by the weights and biases. This time the combining happens *across feature dimensions*, vertically, and each feature grows from 3 numbers to 4. Note that every position goes through the same weight matrix. That is what "position-wise" means. 5. ReLU We cross out the negatives. They become zeros. 6. Second layer Let us bring it back down: 4 dimensions to 3. The output feeds the next block, which has a completely separate set of parameters, and the whole thing runs again. You have just calculated a transformer block by hand. ✍️ The takeaway: the two parts are doing two different jobs, and neither one alone is enough. Attention mixes *across positions*, so a feature can see its neighbours. The FFN mixes *across feature dimensions*, so each position can think about itself. Horizontal, then vertical. Then that pattern repeats N times, each block with its own separate set of weights. That is the Nx from the list up top, and that is what makes the transformer run. 💾 Save this post! #AIbyHand #Transformers #DeepLearning

Tom Yeh

26,065 views • 1 month ago

Transformer by hand ✍️ ~ 6 steps walkthrough below Studying the Transformer architecture is like opening up the engine hood of your car. So many unknown parts: embeddings, positional encoding, attention weighting, self-attention, cross-attention, multi-head attention, layer norm, skip connections, softmax, linear, Nx, shifted right, query, key, value, masking. But which of those actually make the car run? I think at the core, there are two most essential components: Attention weighting and the feed-forward network. Everything else is an enhancement to make it run faster and longer, which is how we got from a car to a truck, and to the word "large" in large language model. So I drew and calculated those two parts entirely by hand. Goal: push five features through one transformer block, filling in every cell yourself. 1. Given Five positions of input features, arriving from the previous block. 2. Attention matrix Let us feed all five features to a query-key module (QK) and read back an attention weight matrix, A. The details of that module are a post of their own. 3. Attention weighting We multiply the input features by A to get the attention weighted features, Z. Still five positions. The effect is to combine features *across positions*, horizontally: X1 becomes X1 + X2, X2 becomes X2 + X3, and so on. 4. First layer Let us feed all five weighted features into the first layer of the FFN. Multiply by the weights and biases. This time the combining happens *across feature dimensions*, vertically, and each feature grows from 3 numbers to 4. Note that every position goes through the same weight matrix. That is what "position-wise" means. 5. ReLU We cross out the negatives. They become zeros. 6. Second layer Let us bring it back down: 4 dimensions to 3. The output feeds the next block, which has a completely separate set of parameters, and the whole thing runs again. You have just calculated a transformer block by hand. ✍️ The takeaway: the two parts are doing two different jobs, and neither one alone is enough. Attention mixes across positions, so a feature can see its neighbours. The FFN mixes across feature dimensions, so each position can think about itself. Horizontal, then vertical. Then that pattern repeats N times, each block with its own separate set of weights. That is the Nx from the list up top, and that is what makes the transformer run. 💾 Save this post! #AIbyHand #Transformers #DeepLearning
0:12

Transformer by hand ✍️ ~ 6 steps walkthrough below Studying the Transformer architecture is like opening up the engine hood of your car. So many unknown parts: embeddings, positional encoding, attention weighting, self-attention, cross-attention, multi-head attention, layer norm, skip connections, softmax, linear, Nx, shifted right, query, key, value, masking. But which of those actually make the car run? I think at the core, there are two most essential components: Attention weighting and the feed-forward network. Everything else is an enhancement to make it run faster and longer, which is how we got from a car to a truck, and to the word "large" in large language model. So I drew and calculated those two parts entirely by hand. Goal: push five features through one transformer block, filling in every cell yourself. 1. Given Five positions of input features, arriving from the previous block. 2. Attention matrix Let us feed all five features to a query-key module (QK) and read back an attention weight matrix, A. The details of that module are a post of their own. 3. Attention weighting We multiply the input features by A to get the attention weighted features, Z. Still five positions. The effect is to combine features *across positions*, horizontally: X1 becomes X1 + X2, X2 becomes X2 + X3, and so on. 4. First layer Let us feed all five weighted features into the first layer of the FFN. Multiply by the weights and biases. This time the combining happens *across feature dimensions*, vertically, and each feature grows from 3 numbers to 4. Note that every position goes through the same weight matrix. That is what "position-wise" means. 5. ReLU We cross out the negatives. They become zeros. 6. Second layer Let us bring it back down: 4 dimensions to 3. The output feeds the next block, which has a completely separate set of parameters, and the whole thing runs again. You have just calculated a transformer block by hand. ✍️ The takeaway: the two parts are doing two different jobs, and neither one alone is enough. Attention mixes across positions, so a feature can see its neighbours. The FFN mixes across feature dimensions, so each position can think about itself. Horizontal, then vertical. Then that pattern repeats N times, each block with its own separate set of weights. That is the Nx from the list up top, and that is what makes the transformer run. 💾 Save this post! #AIbyHand #Transformers #DeepLearning

Tom Yeh

17,396 views • 2 days ago

a moonshot engineer leaked the benchmark anthropic, openai and xai all buried the same week: kimi k3 beat opus 5, gpt-5.6 and grok 4.6 at $0.94 a task. stop paying anthropic $200 a month for opus 5 and openai $200 for gpt-5.6 when kimi does the same work for $8 the leak showed kimi k3 winning 9 of 12 categories against opus 5, gpt-5.6 and grok 4.6. within 48 hours all three labs quietly pushed pricing pages and one very specific comparison chart off their sites. nobody announced anything. they just deleted, which tells you everything the four numbers they scrubbed: cost per task · $0.94 vs $1.80 -> opus 5 charges $1.80 to finish one task. gpt-5.6 $1.04. grok 4.6 $0.61. kimi k3 $0.94 and it landed 487 of 500 clean -> anthropic is billing you double for a model that lost the benchmark it paid to promote the weights · free, sitting on huggingface right now -> the entire model is a public download. pull it, keep it, run it forever, nobody can switch it off -> a model you can hold cannot be rented at $200 a month. that single fact is what three labs deleted a chart over the switch · one line of bash -> moonshot ships an anthropic-compatible endpoint. one env variable and claude code points at kimi -> same cli, same keybindings, same /model. you change a url, opus 5 never knows it lost the seat the bill · $400 down to $8 -> opus 5 max plus gpt-5.6 pro is $400 a month. kimi runs the same daily work for $8 metered -> that is a 98% cut for output that beat both of them 9 categories to 3 here is the part they will fight me on: the frontier tax died the week this leaked and all three labs know it. once the weights are public the price has a ceiling, because anyone can serve the same model. anthropic, openai and xai are charging 2025 prices on a lead that ended in a benchmark they deleted instead of answered drop your $400/mo ai stack to $8. the run above is kimi k3 finishing the task opus 5 bills $1.80 for. the full breakdown is in the article below
0:19

a moonshot engineer leaked the benchmark anthropic, openai and xai all buried the same week: kimi k3 beat opus 5, gpt-5.6 and grok 4.6 at $0.94 a task. stop paying anthropic $200 a month for opus 5 and openai $200 for gpt-5.6 when kimi does the same work for $8 the leak showed kimi k3 winning 9 of 12 categories against opus 5, gpt-5.6 and grok 4.6. within 48 hours all three labs quietly pushed pricing pages and one very specific comparison chart off their sites. nobody announced anything. they just deleted, which tells you everything the four numbers they scrubbed: cost per task · $0.94 vs $1.80 -> opus 5 charges $1.80 to finish one task. gpt-5.6 $1.04. grok 4.6 $0.61. kimi k3 $0.94 and it landed 487 of 500 clean -> anthropic is billing you double for a model that lost the benchmark it paid to promote the weights · free, sitting on huggingface right now -> the entire model is a public download. pull it, keep it, run it forever, nobody can switch it off -> a model you can hold cannot be rented at $200 a month. that single fact is what three labs deleted a chart over the switch · one line of bash -> moonshot ships an anthropic-compatible endpoint. one env variable and claude code points at kimi -> same cli, same keybindings, same /model. you change a url, opus 5 never knows it lost the seat the bill · $400 down to $8 -> opus 5 max plus gpt-5.6 pro is $400 a month. kimi runs the same daily work for $8 metered -> that is a 98% cut for output that beat both of them 9 categories to 3 here is the part they will fight me on: the frontier tax died the week this leaked and all three labs know it. once the weights are public the price has a ceiling, because anyone can serve the same model. anthropic, openai and xai are charging 2025 prices on a lead that ended in a benchmark they deleted instead of answered drop your $400/mo ai stack to $8. the run above is kimi k3 finishing the task opus 5 bills $1.80 for. the full breakdown is in the article below

starmex

30,987 views • 3 days ago

This is absolutely amazing with Flux 3’s split-screen rendering. More power to storytelling! Upscaled with Topaz Astra. Prompt : Split-screen video. Two equal vertical halves. Both halves show the SAME event, at the SAME time, frame-synchronized, filmed by two different cameras. SCENE: A quiet street, daytime. A long tall hedge wall runs along the sidewalk, too tall to see over. There is one narrow gap in the hedge ahead. A woman in a red jacket walks alone along the sidewalk on the near side of the hedge. On the FAR side of the hedge is an open grass field, where a large golden dog runs. LEFT HALF — CAMERA A: ground-level tracking shot on the sidewalk, following the woman from behind at shoulder height. IMPORTANT: from this camera, only the woman, the sidewalk, and the hedge wall are visible. The field, the dog, and anything behind the hedge are NEVER visible in the left half. The street looks calm and empty. RIGHT HALF — CAMERA B: aerial top-down drone shot, directly above, moving with the woman. This view shows BOTH sides of the hedge at once: the woman on the sidewalk, the hedge as a thin green line, and the golden dog sprinting across the field on the other side, on a converging path toward the gap in the hedge. ACTION — identical timing in both halves: 0–8s: The woman walks calmly. LEFT: peaceful, nothing unusual. RIGHT: the dog races closer and closer to the gap, its path and the woman's path clearly about to meet. 8–11s: The woman reaches the gap. The dog bursts through it. LEFT: the dog appears suddenly from nowhere, a total surprise; the woman flinches. RIGHT: the meeting looks perfectly predictable, two paths joining. 11–15s: The dog jumps up joyfully; it is her own dog greeting her. She laughs, kneels, and hugs it. Both halves show this ending. RULES: Same woman, same dog, same timing in both halves. The dog is visible ONLY in the right half until 8s. No cuts, no other people, no cars.
0:15

This is absolutely amazing with Flux 3’s split-screen rendering. More power to storytelling! Upscaled with Topaz Astra. Prompt : Split-screen video. Two equal vertical halves. Both halves show the SAME event, at the SAME time, frame-synchronized, filmed by two different cameras. SCENE: A quiet street, daytime. A long tall hedge wall runs along the sidewalk, too tall to see over. There is one narrow gap in the hedge ahead. A woman in a red jacket walks alone along the sidewalk on the near side of the hedge. On the FAR side of the hedge is an open grass field, where a large golden dog runs. LEFT HALF — CAMERA A: ground-level tracking shot on the sidewalk, following the woman from behind at shoulder height. IMPORTANT: from this camera, only the woman, the sidewalk, and the hedge wall are visible. The field, the dog, and anything behind the hedge are NEVER visible in the left half. The street looks calm and empty. RIGHT HALF — CAMERA B: aerial top-down drone shot, directly above, moving with the woman. This view shows BOTH sides of the hedge at once: the woman on the sidewalk, the hedge as a thin green line, and the golden dog sprinting across the field on the other side, on a converging path toward the gap in the hedge. ACTION — identical timing in both halves: 0–8s: The woman walks calmly. LEFT: peaceful, nothing unusual. RIGHT: the dog races closer and closer to the gap, its path and the woman's path clearly about to meet. 8–11s: The woman reaches the gap. The dog bursts through it. LEFT: the dog appears suddenly from nowhere, a total surprise; the woman flinches. RIGHT: the meeting looks perfectly predictable, two paths joining. 11–15s: The dog jumps up joyfully; it is her own dog greeting her. She laughs, kneels, and hugs it. Both halves show this ending. RULES: Same woman, same dog, same timing in both halves. The dog is visible ONLY in the right half until 8s. No cuts, no other people, no cars.

Umesh

35,059 views • 29 days ago

Vector Database by Hand ✍️ Vector databases are revolutionizing how we search and analyze complex data. They have become the backbone of Retrieval Augmented Generation (#RAG). How do vector databases work? [1] Given ↳ A dataset of three sentences, each has 3 words (or tokens) ↳ In practice, a dataset may contain millions or billions of sentences. The max number of tokens may be tens of thousands (e.g., 32,768 mistral-7b). Process "how are you" [2] 🟨 Word Embeddings ↳ For each word, look up corresponding word embedding vector from a table of 22 vectors, where 22 is the vocabulary size. ↳ In practice, the vocabulary size can be tens of thousands. The word embedding dimensions are in the thousands (e.g., 1024, 4096) [3] 🟩 Encoding ↳ Feed the sequence of word embeddings to an encoder to obtain a sequence of feature vectors, one per word. ↳ Here, the encoder is a simple one layer perceptron (linear layer + ReLU) ↳ In practice, the encoder is a transformer or one of its many variants. [4] 🟩 Mean Pooling ↳ Merge the sequence of feature vectors into a single vector using "mean pooling" which is to average across the columns. ↳ The result is a single vector. We often call it "text embeddings" or "sentence embeddings." ↳ Other pooling techniques are possible, such as CLS. But mean pooling is the most common. [5] 🟦 Indexing ↳ Reduce the dimensions of the text embedding vector by a projection matrix. The reduction rate is 50% (4->2). ↳ In practice, the values in this projection matrix is much more random. ↳ The purpose is similar to that of hashing, which is to obtain a short representation to allow faster comparison and retrieval. ↳ The resulting dimension-reduced index vector is saved in the vector storage. [6] Process "who are you" ↳ Repeat [2]-[5] [7] Process "who am I" ↳ Repeat [2]-[5] Now we have indexed our dataset in the vector database. [8] 🟥 Query: "am I you" ↳ Repeat [2]-[5] ↳ The result is a 2-d query vector. [9] 🟥 Dot Products ↳ Take dot product between the query vector and database vectors. They are all 2-d. ↳ The purpose is to use dot product to estimate similarity. ↳ By transposing the query vector, this step becomes a matrix multiplication. [10] 🟥 Nearest Neighbor ↳ Find the largest dot product by linear scan. ↳ The sentence with the highest dot product is "who am I" ↳ In practice, because scanning billions of vectors is slow, we use an Approximate Nearest Neighbor (ANN) algorithm like the Hierarchical Navigable Small Worlds (HNSW).
0:17

Vector Database by Hand ✍️ Vector databases are revolutionizing how we search and analyze complex data. They have become the backbone of Retrieval Augmented Generation (#RAG). How do vector databases work? [1] Given ↳ A dataset of three sentences, each has 3 words (or tokens) ↳ In practice, a dataset may contain millions or billions of sentences. The max number of tokens may be tens of thousands (e.g., 32,768 mistral-7b). Process "how are you" [2] 🟨 Word Embeddings ↳ For each word, look up corresponding word embedding vector from a table of 22 vectors, where 22 is the vocabulary size. ↳ In practice, the vocabulary size can be tens of thousands. The word embedding dimensions are in the thousands (e.g., 1024, 4096) [3] 🟩 Encoding ↳ Feed the sequence of word embeddings to an encoder to obtain a sequence of feature vectors, one per word. ↳ Here, the encoder is a simple one layer perceptron (linear layer + ReLU) ↳ In practice, the encoder is a transformer or one of its many variants. [4] 🟩 Mean Pooling ↳ Merge the sequence of feature vectors into a single vector using "mean pooling" which is to average across the columns. ↳ The result is a single vector. We often call it "text embeddings" or "sentence embeddings." ↳ Other pooling techniques are possible, such as CLS. But mean pooling is the most common. [5] 🟦 Indexing ↳ Reduce the dimensions of the text embedding vector by a projection matrix. The reduction rate is 50% (4->2). ↳ In practice, the values in this projection matrix is much more random. ↳ The purpose is similar to that of hashing, which is to obtain a short representation to allow faster comparison and retrieval. ↳ The resulting dimension-reduced index vector is saved in the vector storage. [6] Process "who are you" ↳ Repeat [2]-[5] [7] Process "who am I" ↳ Repeat [2]-[5] Now we have indexed our dataset in the vector database. [8] 🟥 Query: "am I you" ↳ Repeat [2]-[5] ↳ The result is a 2-d query vector. [9] 🟥 Dot Products ↳ Take dot product between the query vector and database vectors. They are all 2-d. ↳ The purpose is to use dot product to estimate similarity. ↳ By transposing the query vector, this step becomes a matrix multiplication. [10] 🟥 Nearest Neighbor ↳ Find the largest dot product by linear scan. ↳ The sentence with the highest dot product is "who am I" ↳ In practice, because scanning billions of vectors is slow, we use an Approximate Nearest Neighbor (ANN) algorithm like the Hierarchical Navigable Small Worlds (HNSW).

Tom Yeh

192,022 views • 2 years ago

you can legally steal any viral ugc, rewrite it for your product, and generate it with AI UGC. people are generating photorealistic AI actors that hold real products in their hands and reading scripts that convert like real UGC. Vodafone already ran a full TikTok campaign with AI influencers. same engagement, same conversions, fraction of the cost here's the ugc system anyone can run today: - find a product that hits an emotional pain point (aging skin, insecurity, frustration). one retinol serum video has 8M views and 300k likes with comments begging for the link - download the viral video, transcribe it with Gemini, then feed the transcript + your product + your audience into Claude. it spits out 3 scripts that sound like a friend talking, not an ad - build your actor in any ai ugc platform: filter by age, location, accessories, or generate one from a prompt. nano banana places your exact product in her hand with real lighting and shadows - create two versions of the same actor. no product for the problem hook, product in hand for the solution. that switch is what makes it feel like a real story - use the scenes feature to generate actual application footage: fingertips dabbing serum, circular motions, natural hand movement. not just a talking head - a UGC creator charges $200 to $600 for this exact video. it takes 2 minutes and you can make 10 variations tomorrow the brands winning right now aren't the ones with the biggest budgets. they're the ones testing fastest. reply "ugc" + RT and i'll send you the full video so you can build this too.
0:18

you can legally steal any viral ugc, rewrite it for your product, and generate it with AI UGC. people are generating photorealistic AI actors that hold real products in their hands and reading scripts that convert like real UGC. Vodafone already ran a full TikTok campaign with AI influencers. same engagement, same conversions, fraction of the cost here's the ugc system anyone can run today: - find a product that hits an emotional pain point (aging skin, insecurity, frustration). one retinol serum video has 8M views and 300k likes with comments begging for the link - download the viral video, transcribe it with Gemini, then feed the transcript + your product + your audience into Claude. it spits out 3 scripts that sound like a friend talking, not an ad - build your actor in any ai ugc platform: filter by age, location, accessories, or generate one from a prompt. nano banana places your exact product in her hand with real lighting and shadows - create two versions of the same actor. no product for the problem hook, product in hand for the solution. that switch is what makes it feel like a real story - use the scenes feature to generate actual application footage: fingertips dabbing serum, circular motions, natural hand movement. not just a talking head - a UGC creator charges $200 to $600 for this exact video. it takes 2 minutes and you can make 10 variations tomorrow the brands winning right now aren't the ones with the biggest budgets. they're the ones testing fastest. reply "ugc" + RT and i'll send you the full video so you can build this too.

Sulfur

17,525 views • 1 month 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 • 13 days ago

Walter is 79 years old. Came into our shelter last Thursday afternoon. Staff immediately started showing him the younger dogs. A two-year-old Labrador. A playful Beagle mix. A one-year-old Retriever. Walter shook his head at every single one. Our volunteer asked him: "What kind of dog are you looking for?" Walter said: "I want the dog nobody else wants. The one people walk past because they think he doesn't have enough time left." We brought Walter to our senior section. Introduced him to Max. Twelve-year-old Shepherd mix. Surrendered three months ago when his previous owner moved into assisted living. Walter knelt down slowly in front of Max's kennel. Took him a minute to get down there. Both of them just looked at each other. Gray hair. Slow movements. Tired eyes. Walter said quietly: "We're the same, you and me. Everyone thinks our best days are behind us. But I've still got love to give. I bet you do too." Signed the papers. Took Max home that day. They move at the same pace. They understand each other. Walter called us two days later and said: "Max is perfect. We take naps together. We're both exactly what the other needed." Age isn't a limitation. It's a language.
0:15

Walter is 79 years old. Came into our shelter last Thursday afternoon. Staff immediately started showing him the younger dogs. A two-year-old Labrador. A playful Beagle mix. A one-year-old Retriever. Walter shook his head at every single one. Our volunteer asked him: "What kind of dog are you looking for?" Walter said: "I want the dog nobody else wants. The one people walk past because they think he doesn't have enough time left." We brought Walter to our senior section. Introduced him to Max. Twelve-year-old Shepherd mix. Surrendered three months ago when his previous owner moved into assisted living. Walter knelt down slowly in front of Max's kennel. Took him a minute to get down there. Both of them just looked at each other. Gray hair. Slow movements. Tired eyes. Walter said quietly: "We're the same, you and me. Everyone thinks our best days are behind us. But I've still got love to give. I bet you do too." Signed the papers. Took Max home that day. They move at the same pace. They understand each other. Walter called us two days later and said: "Max is perfect. We take naps together. We're both exactly what the other needed." Age isn't a limitation. It's a language.

Gabriele Corno

86,132 views • 9 days ago

BREAKING 🚨Google’s quantum chip didn’t prove we live in a multiverse. It just proved the universe is one beautifully connected sea.🧨 Google’s latest quantum chip solved a problem in five minutes that would take a classical supercomputer 10 septillion years. Some physicists are calling it proof of a multiverse — the idea that every possible outcome branches into its own reality, so the chip is somehow “sampling” answers from parallel universes. Uniphics shows there is no need for any multiverse. Everything is made of spinning Gyrotrons whose waves propagate through one single ξM-field sea of unbound energy that fills all space. When the quantum chip sets up its qubits, those Gyrotrons create vast networks of perfectly coherent spin waves. Because the waves interfere across the entire sea at once, the chip can explore enormous numbers of possibilities simultaneously — not by jumping into other universes, but by letting the single connected field do what it always does: keep perfect harmony across its entire volume. The speed-up comes from the natural parallelism of spin-wave interference plus local time-flow variations (t_flow = k / E_d,total) that let dense regions of the chip run on slightly different clocks, giving the appearance of massive parallel computation without ever leaving our one deterministic universe. The same three pillars that explain gravity as a push and the cyclic cosmos also turn quantum computing into simple, single-universe physics. The universe isn’t splitting into trillions of realities every time a chip runs. It’s simply one sea singing in perfect harmony — and Google just learned a new note. How soon will quantum computing explode when we stop inventing multiverses and start engineering the single connected sea? A Theory of Everything should be able to answer everything. Uniphics Explained Simply PDF: Chapters 1–10 free:
0:10

BREAKING 🚨Google’s quantum chip didn’t prove we live in a multiverse. It just proved the universe is one beautifully connected sea.🧨 Google’s latest quantum chip solved a problem in five minutes that would take a classical supercomputer 10 septillion years. Some physicists are calling it proof of a multiverse — the idea that every possible outcome branches into its own reality, so the chip is somehow “sampling” answers from parallel universes. Uniphics shows there is no need for any multiverse. Everything is made of spinning Gyrotrons whose waves propagate through one single ξM-field sea of unbound energy that fills all space. When the quantum chip sets up its qubits, those Gyrotrons create vast networks of perfectly coherent spin waves. Because the waves interfere across the entire sea at once, the chip can explore enormous numbers of possibilities simultaneously — not by jumping into other universes, but by letting the single connected field do what it always does: keep perfect harmony across its entire volume. The speed-up comes from the natural parallelism of spin-wave interference plus local time-flow variations (t_flow = k / E_d,total) that let dense regions of the chip run on slightly different clocks, giving the appearance of massive parallel computation without ever leaving our one deterministic universe. The same three pillars that explain gravity as a push and the cyclic cosmos also turn quantum computing into simple, single-universe physics. The universe isn’t splitting into trillions of realities every time a chip runs. It’s simply one sea singing in perfect harmony — and Google just learned a new note. How soon will quantum computing explode when we stop inventing multiverses and start engineering the single connected sea? A Theory of Everything should be able to answer everything. Uniphics Explained Simply PDF: Chapters 1–10 free:

Paul Maley

46,801 views • 5 months ago

They did not take cursive from the schools because children no longer needed it. They took it because of what it was quietly building in them. Consider what the exercise actually is. A child, six years old, is handed a pen and asked to draw a single unbroken line that becomes a word. The wrist must float. The fingers must hold a living pressure, never quite the same twice, always correcting. The eye must follow the ink forward and trust the hand to finish what it has begun. There is no lifting, no stopping, no starting over mid-word. The loop must close. The ascender must rise and return. The sentence must travel from one margin to the other as a single continuous gesture, and at the end of it the hand must still be steady. Twelve years of this. Every day. Ten thousand small acts of sustained, self-correcting attention, carried out below the level of conscious thought, until the motion belongs to the body and the body belongs to the motion. This is not penmanship. It is the slow construction of an interior form. The hand that has learned to carry a line without breaking it is the hand of a mind that has learned to carry a thought without breaking it. The two are not metaphors for one another. They are the same faculty, trained in the same child, by the same daily discipline. Continuity of the stroke becomes continuity of the reasoning. The patience of the loop becomes the patience of the argument. The commitment to finish a word one has started becomes the commitment to finish a sentence, a paragraph, a life's idea, without reaching for the nearest distraction halfway through. Print is a different creature entirely. Print lifts. Print stops. Print assembles a word out of separate, stamped, interchangeable pieces, each one beginning and ending in isolation. A mind raised only on print learns to think the way print is made, in discrete tokens, in replaceable units, in fragments that can be recombined by any outside hand without the owner noticing the substitution. It is precisely the shape of thought a language model produces. It is precisely the shape of thought a language model can steer. Cursive is kata. This is the whole of it. A form repeated daily, for years, not for the sake of the form but for what the repetition lays down in the practitioner beneath the form. The swordsman does not train kata so that one day he may fight in kata. He trains it so that when the moment comes and there is no time to think, the movement is already inside him, older and deeper than thought, and it rises on its own. Cursive was the kata of the literate mind, the daily quiet drilling of continuity, of patience, of a line held steady under the long pressure of its own length. And the signature it produced at the end, that small flourished mark unique to a single human being on earth, was only the outward proof of an inward form no machine and no other hand could ever reproduce. Take the kata away and the practitioner is left with vocabulary in place of faculty. He can recognise a whole thought when he encounters one. He cannot carry one himself. He can admire a finished argument. He cannot sustain one long enough to close its loop. He begins books he does not finish, sentences he does not end, ideas he abandons the moment the screen in his palm offers him a brighter one. And when the machine begins feeding him tokens in the exact shape his schooling taught him to receive, he meets it with no interior resistance at all, because no interior form was ever built in him to push back with. They removed it quietly, across a generation, and they removed it in the last years before the machines arrived. Twelve years of daily practice in unbroken, embodied, self-authored thought, gone from the curriculum of almost every child in the Western world, just as the instruments designed to complete their sentences for them came online. The hand forgets. The mind, having never been taught the kata, forgets a thing it never knew it had. That is what cursive was. That is what was taken. And that is why the thought of anyone who still writes by hand, in long unlifted lines, remains, quietly, stubbornly, and without their ever needing to announce it, their own. Now the question stands open. What else has been banned, phased out, quietly retired from the curriculum and from common life over these same decades, under the same soft excuses? Mental arithmetic. Memorisation of poetry. Latin. Logic as a formal subject. Map reading. Knot work. The keeping of a commonplace book. The reading aloud of long passages in class. Singing in parts. What was each of those actually building in the child, beneath the surface of the lesson, and whose interest was served by its disappearance?
0:10

They did not take cursive from the schools because children no longer needed it. They took it because of what it was quietly building in them. Consider what the exercise actually is. A child, six years old, is handed a pen and asked to draw a single unbroken line that becomes a word. The wrist must float. The fingers must hold a living pressure, never quite the same twice, always correcting. The eye must follow the ink forward and trust the hand to finish what it has begun. There is no lifting, no stopping, no starting over mid-word. The loop must close. The ascender must rise and return. The sentence must travel from one margin to the other as a single continuous gesture, and at the end of it the hand must still be steady. Twelve years of this. Every day. Ten thousand small acts of sustained, self-correcting attention, carried out below the level of conscious thought, until the motion belongs to the body and the body belongs to the motion. This is not penmanship. It is the slow construction of an interior form. The hand that has learned to carry a line without breaking it is the hand of a mind that has learned to carry a thought without breaking it. The two are not metaphors for one another. They are the same faculty, trained in the same child, by the same daily discipline. Continuity of the stroke becomes continuity of the reasoning. The patience of the loop becomes the patience of the argument. The commitment to finish a word one has started becomes the commitment to finish a sentence, a paragraph, a life's idea, without reaching for the nearest distraction halfway through. Print is a different creature entirely. Print lifts. Print stops. Print assembles a word out of separate, stamped, interchangeable pieces, each one beginning and ending in isolation. A mind raised only on print learns to think the way print is made, in discrete tokens, in replaceable units, in fragments that can be recombined by any outside hand without the owner noticing the substitution. It is precisely the shape of thought a language model produces. It is precisely the shape of thought a language model can steer. Cursive is kata. This is the whole of it. A form repeated daily, for years, not for the sake of the form but for what the repetition lays down in the practitioner beneath the form. The swordsman does not train kata so that one day he may fight in kata. He trains it so that when the moment comes and there is no time to think, the movement is already inside him, older and deeper than thought, and it rises on its own. Cursive was the kata of the literate mind, the daily quiet drilling of continuity, of patience, of a line held steady under the long pressure of its own length. And the signature it produced at the end, that small flourished mark unique to a single human being on earth, was only the outward proof of an inward form no machine and no other hand could ever reproduce. Take the kata away and the practitioner is left with vocabulary in place of faculty. He can recognise a whole thought when he encounters one. He cannot carry one himself. He can admire a finished argument. He cannot sustain one long enough to close its loop. He begins books he does not finish, sentences he does not end, ideas he abandons the moment the screen in his palm offers him a brighter one. And when the machine begins feeding him tokens in the exact shape his schooling taught him to receive, he meets it with no interior resistance at all, because no interior form was ever built in him to push back with. They removed it quietly, across a generation, and they removed it in the last years before the machines arrived. Twelve years of daily practice in unbroken, embodied, self-authored thought, gone from the curriculum of almost every child in the Western world, just as the instruments designed to complete their sentences for them came online. The hand forgets. The mind, having never been taught the kata, forgets a thing it never knew it had. That is what cursive was. That is what was taken. And that is why the thought of anyone who still writes by hand, in long unlifted lines, remains, quietly, stubbornly, and without their ever needing to announce it, their own. Now the question stands open. What else has been banned, phased out, quietly retired from the curriculum and from common life over these same decades, under the same soft excuses? Mental arithmetic. Memorisation of poetry. Latin. Logic as a formal subject. Map reading. Knot work. The keeping of a commonplace book. The reading aloud of long passages in class. Singing in parts. What was each of those actually building in the child, beneath the surface of the lesson, and whose interest was served by its disappearance?

SiriusB

443,560 views • 4 months ago

This guy built a visual scanner that reads 468 points on his face and 42 points on his hands from a regular webcam and turns them into a cloud of thousands of particles right between his palms. Inside, MediaPipe and TouchDesigner are linked: the first captures hands and face from the webcam with high accuracy, the second turns those coordinates into a live plane and feeds it into a POP system that instantly generates a swarm of particles in the shape of a head. No studio, no render farmer, no VR headset. Just a laptop, a webcam, and 1 TouchDesigner session. And traditional VJ studios keep teams of 5 people on a setup with lighting, custom hardware, and commercial plugins, while his expenses are only a TouchDesigner subscription and a regular USB camera. One laptop runs MediaPipe and TouchDesigner simultaneously, holds the camera stream at 60 FPS without drops, and in parallel processes 468 face points + 21 points on each hand. The camera captures frame after frame, MediaPipe in real time sends TouchDesigner the finger coordinates and face geometry, and the POP operator inside the engine translates those numbers into thousands of particle points with colors from bright pink to gold. This setup immediately defines the role of the tool and the limits of its autonomy. It knows where the fingertips are at every moment of the frame. It knows how to read the face geometry at any angle to the camera. It knows how to draw a swarm of particles between them with the right color and contour. → MediaPipe pulls 468 points from the face and 21 points from each hand, 60 times per second → TouchDesigner receives those coordinates, builds a virtual rectangle between the fingertips, and feeds it into the POP system → POP generates thousands of particle points in the shape of a head, coloring them in a gradient from bright pink to gold → The HUD layer adds green corners and a blue neon frame, styling the image like an AR interface → All layers assemble into 1 real-time frame that projects back onto the video in the camera window → The final image is recorded to a file or broadcast to a projector for a live installation And only when the guy spreads his hands wider does the plane between the palms stretch; brings them together, it narrows. Otherwise the system runs on its own. And when he moves from his home room to a concert hall, the same laptop with the same webcam launches the same TouchDesigner session in just 5 minutes, without reconfiguration, without a new team, and without a single line of new code. In his work setup there is no studio of his own and no team for assembly. On the desk sits a laptop with a webcam, on top run MediaPipe and TouchDesigner with POP operators, and the same setup through a USB camera moves to any concert without a new configuration. Out of everything I have seen this year, this is the cleanest Creative Coding setup on 1 laptop: 0 render farms, 0 studio lighting, and between them 3 libraries, thousands of particle points, and 1 webcam.
0:10

This guy built a visual scanner that reads 468 points on his face and 42 points on his hands from a regular webcam and turns them into a cloud of thousands of particles right between his palms. Inside, MediaPipe and TouchDesigner are linked: the first captures hands and face from the webcam with high accuracy, the second turns those coordinates into a live plane and feeds it into a POP system that instantly generates a swarm of particles in the shape of a head. No studio, no render farmer, no VR headset. Just a laptop, a webcam, and 1 TouchDesigner session. And traditional VJ studios keep teams of 5 people on a setup with lighting, custom hardware, and commercial plugins, while his expenses are only a TouchDesigner subscription and a regular USB camera. One laptop runs MediaPipe and TouchDesigner simultaneously, holds the camera stream at 60 FPS without drops, and in parallel processes 468 face points + 21 points on each hand. The camera captures frame after frame, MediaPipe in real time sends TouchDesigner the finger coordinates and face geometry, and the POP operator inside the engine translates those numbers into thousands of particle points with colors from bright pink to gold. This setup immediately defines the role of the tool and the limits of its autonomy. It knows where the fingertips are at every moment of the frame. It knows how to read the face geometry at any angle to the camera. It knows how to draw a swarm of particles between them with the right color and contour. → MediaPipe pulls 468 points from the face and 21 points from each hand, 60 times per second → TouchDesigner receives those coordinates, builds a virtual rectangle between the fingertips, and feeds it into the POP system → POP generates thousands of particle points in the shape of a head, coloring them in a gradient from bright pink to gold → The HUD layer adds green corners and a blue neon frame, styling the image like an AR interface → All layers assemble into 1 real-time frame that projects back onto the video in the camera window → The final image is recorded to a file or broadcast to a projector for a live installation And only when the guy spreads his hands wider does the plane between the palms stretch; brings them together, it narrows. Otherwise the system runs on its own. And when he moves from his home room to a concert hall, the same laptop with the same webcam launches the same TouchDesigner session in just 5 minutes, without reconfiguration, without a new team, and without a single line of new code. In his work setup there is no studio of his own and no team for assembly. On the desk sits a laptop with a webcam, on top run MediaPipe and TouchDesigner with POP operators, and the same setup through a USB camera moves to any concert without a new configuration. Out of everything I have seen this year, this is the cleanest Creative Coding setup on 1 laptop: 0 render farms, 0 studio lighting, and between them 3 libraries, thousands of particle points, and 1 webcam.

Blaze

38,242 views • 3 months 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

719,133 views • 14 days ago

whoever leaked this has bigger balls than sense Google Research and MIT ran the same agent jobs 260 different ways for Nature last month: they held the prompts, the tools and the compute budget identical and moved nothing but the wiring between the agents, and the same work swung from 70% worse than a single agent to 80.8% better, averaging out at 0.0% i ran my own single agent against the task list first and it cleared 6 of 10 alone, already past the line where a crew starts subtracting this is Graph Engineering, the layer that decides whether a crew is worth 80% more or 70% less, and it installs into the agent you already pay for: - score your solo agent on the real task first: above roughly 45% success that study predicts zero to negative returns from any crew you put around it - under that line, put one supervisor over the fan out: crews with no correction step amplified their own errors to 17.2x the single agent rate, supervised aggregation held it to 4.4x - give every worker one output and let none of them read a peer's draft, so a wrong step reaches the supervisor instead of four other agents - run the comparison again after every model upgrade, because a better model raises your baseline and a higher baseline is what makes a crew stop paying - keep the single agent alive as the control, the only number that says the wiring is earning its calls turns out the shape does not travel: the biggest win came off a finance task under one supervisor and the worst collapse off a planning task with independent agents my position, and it is the arguable one: a crew is a bet on your own diagram, and the model you pick moves that bet less than one arrow does bookmark this, the three moves that draw those arrows before you pay for one extra call are in the post below ↓
0:15

whoever leaked this has bigger balls than sense Google Research and MIT ran the same agent jobs 260 different ways for Nature last month: they held the prompts, the tools and the compute budget identical and moved nothing but the wiring between the agents, and the same work swung from 70% worse than a single agent to 80.8% better, averaging out at 0.0% i ran my own single agent against the task list first and it cleared 6 of 10 alone, already past the line where a crew starts subtracting this is Graph Engineering, the layer that decides whether a crew is worth 80% more or 70% less, and it installs into the agent you already pay for: - score your solo agent on the real task first: above roughly 45% success that study predicts zero to negative returns from any crew you put around it - under that line, put one supervisor over the fan out: crews with no correction step amplified their own errors to 17.2x the single agent rate, supervised aggregation held it to 4.4x - give every worker one output and let none of them read a peer's draft, so a wrong step reaches the supervisor instead of four other agents - run the comparison again after every model upgrade, because a better model raises your baseline and a higher baseline is what makes a crew stop paying - keep the single agent alive as the control, the only number that says the wiring is earning its calls turns out the shape does not travel: the biggest win came off a finance task under one supervisor and the worst collapse off a planning task with independent agents my position, and it is the arguable one: a crew is a bet on your own diagram, and the model you pick moves that bet less than one arrow does bookmark this, the three moves that draw those arrows before you pay for one extra call are in the post below ↓

Argona

888,684 views • 11 days ago

Everyone is trying to find a 10x trade on Polymarket. These 5 markets are sitting wide open printing free money. And almost nobody is structuring them correctly. > Here's what I mean: Market 1: Jesus doesn't return before 2027 Market 2: Iran doesn't legalize gay marriage Market 3: US stays in NATO before 2027 Market 4: Aliens not confirmed before 2027 Market 5: Putin stays in power before 2027 [ > Chance any of these happen? Zero. Literally zero. Each one pays 5-20% traded separately. Sounds small so most people move on. That's the most expensive mistake on Polymarket right now. Here's why: PolyParlay doesn't add your returns together. It multiplies them. Every near-certain outcome you stack into one position compounds the payout of every other leg. Real math: > 5 markets at 95% probability traded one by one = +5% collected five times = +25% total > Same 5 combined into one parlay = +1,000%+ in a single position Same capital. Same outcomes. Same near-zero risk. 40x more money collected just by changing the structure. Bot access: I already opened this parlay. Not because I found something nobody knows about. Because I stopped leaving money behind on trades that are already obvious. That's the entire difference between traders who grind and traders who print. Free money markets sitting wide open. Wrong structure costs you 40x every single time. One parlay fixes that permanently.
0:19

Everyone is trying to find a 10x trade on Polymarket. These 5 markets are sitting wide open printing free money. And almost nobody is structuring them correctly. > Here's what I mean: Market 1: Jesus doesn't return before 2027 Market 2: Iran doesn't legalize gay marriage Market 3: US stays in NATO before 2027 Market 4: Aliens not confirmed before 2027 Market 5: Putin stays in power before 2027 [ > Chance any of these happen? Zero. Literally zero. Each one pays 5-20% traded separately. Sounds small so most people move on. That's the most expensive mistake on Polymarket right now. Here's why: PolyParlay doesn't add your returns together. It multiplies them. Every near-certain outcome you stack into one position compounds the payout of every other leg. Real math: > 5 markets at 95% probability traded one by one = +5% collected five times = +25% total > Same 5 combined into one parlay = +1,000%+ in a single position Same capital. Same outcomes. Same near-zero risk. 40x more money collected just by changing the structure. Bot access: I already opened this parlay. Not because I found something nobody knows about. Because I stopped leaving money behind on trades that are already obvious. That's the entire difference between traders who grind and traders who print. Free money markets sitting wide open. Wrong structure costs you 40x every single time. One parlay fixes that permanently.

Atenov int.

11,895 views • 3 months ago

Don't train the model, evolve the harness. I read a brilliant blog post from Hugging Face where they took a frozen open model scoring 0% on a hard legal agent benchmark, left its weights alone, and let an automated loop rewrite only the code around it. That code layer is the harness, the runtime wrapper that feeds the model context, runs its tool calls, and decides when a run ends. By the time the loop finished, the system had essentially matched Sonnet 4.6 on the benchmark's headline metric, at roughly 7x lower cost per task. Zero weights changed. The gain existed because of where the model was failing. The judge only grades files saved in the right place under the exact requested filename, and the model kept doing the legal analysis correctly, then saving it under the wrong name, dropping it in a scratch folder, or never writing it at all. So the 0% was never measuring legal reasoning. It was measuring the harness. Hand-tuning that layer is slow and model-specific, so they automated it. A Claude proposer adds exactly one mechanism per iteration, and an outer loop keeps it only if it clearly beats the current best, so accepted mechanisms compound. What the loop discovered says a lot about where agents actually fail. → The biggest single gain was file handling, not intelligence. An automatic step that lands the deliverable exactly where the judge expects it beat every prompt change, with zero extra model tokens. → Code fixes transferred across models, prompt playbooks did not. The same harness lifted a smaller model from the same family by 14 points, but the tuned prompts hurt a different model family on tasks it could already finish. → The harness mattered more than anything else. Same model, same judge, same tasks, and five different harnesses scored anywhere between 3.5% and 80.1%. The gains do eventually flatten, and the remaining misses look like real capability gaps. At some point the wrapper runs out of tricks and the model has to carry the work. But the lesson holds. A benchmark score measures the model and its harness together, and until the harness is fixed, it's impossible to know which one failed. I highly recommend reading this: I also wrote a deep dive on agent harness engineering a while back, covering the orchestration loop, tools, memory, context management, and everything that turns a stateless LLM into a capable agent. The article is quoted below.
0:18

Don't train the model, evolve the harness. I read a brilliant blog post from Hugging Face where they took a frozen open model scoring 0% on a hard legal agent benchmark, left its weights alone, and let an automated loop rewrite only the code around it. That code layer is the harness, the runtime wrapper that feeds the model context, runs its tool calls, and decides when a run ends. By the time the loop finished, the system had essentially matched Sonnet 4.6 on the benchmark's headline metric, at roughly 7x lower cost per task. Zero weights changed. The gain existed because of where the model was failing. The judge only grades files saved in the right place under the exact requested filename, and the model kept doing the legal analysis correctly, then saving it under the wrong name, dropping it in a scratch folder, or never writing it at all. So the 0% was never measuring legal reasoning. It was measuring the harness. Hand-tuning that layer is slow and model-specific, so they automated it. A Claude proposer adds exactly one mechanism per iteration, and an outer loop keeps it only if it clearly beats the current best, so accepted mechanisms compound. What the loop discovered says a lot about where agents actually fail. → The biggest single gain was file handling, not intelligence. An automatic step that lands the deliverable exactly where the judge expects it beat every prompt change, with zero extra model tokens. → Code fixes transferred across models, prompt playbooks did not. The same harness lifted a smaller model from the same family by 14 points, but the tuned prompts hurt a different model family on tasks it could already finish. → The harness mattered more than anything else. Same model, same judge, same tasks, and five different harnesses scored anywhere between 3.5% and 80.1%. The gains do eventually flatten, and the remaining misses look like real capability gaps. At some point the wrapper runs out of tricks and the model has to carry the work. But the lesson holds. A benchmark score measures the model and its harness together, and until the harness is fixed, it's impossible to know which one failed. I highly recommend reading this: I also wrote a deep dive on agent harness engineering a while back, covering the orchestration loop, tools, memory, context management, and everything that turns a stateless LLM into a capable agent. The article is quoted below.

Akshay 🚀

244,885 views • 1 month ago

🚨 THE BIGGEST BOTTLENECK IN AI ISN'T COMPUTING POWER ANYMORE IT'S MOVING DATA. Instead of laying new cables, Chinese researchers have upgraded existing fiber infrastructure by doing two things at once: Using three wavelength bands (C + L + S) instead of the usual two. Using four cores inside each fiber instead of one. Each core acts like an independent highway, and each band acts like an extra lane on that highway. Together, they’ve reportedly increased transmission capacity per core by nearly 50% and overall data throughput by up to 5×. This matters enormously for AI. Modern AI clusters move terabits of data per second between thousands of GPUs. The biggest bottleneck is often not the chips themselves, but moving data fast enough between them. If you can push 5× more data through the same physical cables, you can train bigger models faster and reduce network congestion. Why this is significant: • It shows multi-core + extended spectrum technology moving from labs into real-world commercial use • The system has already run over 35 km of existing telecom network • It could be especially useful for submarine cables and large-scale data center interconnects • China is also eyeing it for its “Eastern Data, Western Computing” project The deeper implication: We’re reaching the physical limits of how much data we can push through single-core fibers using traditional methods. By combining spatial multiplexing (multiple cores) with spectral multiplexing (more wavelength bands), engineers are finding new ways to keep scaling bandwidth without having to dig up the planet to lay new cables. This kind of breakthrough is quiet but foundational it’s the kind of infrastructure upgrade that will determine how fast AI and cloud computing can actually grow in the coming years. The future of data movement might not require more cables. It might just require smarter ones. How important do you think multi-core and multi-band fiber will be for keeping up with AI’s exploding data demands? Follow for more frontier networking, photonics, and infrastructure technology.
0:10

🚨 THE BIGGEST BOTTLENECK IN AI ISN'T COMPUTING POWER ANYMORE IT'S MOVING DATA. Instead of laying new cables, Chinese researchers have upgraded existing fiber infrastructure by doing two things at once: Using three wavelength bands (C + L + S) instead of the usual two. Using four cores inside each fiber instead of one. Each core acts like an independent highway, and each band acts like an extra lane on that highway. Together, they’ve reportedly increased transmission capacity per core by nearly 50% and overall data throughput by up to 5×. This matters enormously for AI. Modern AI clusters move terabits of data per second between thousands of GPUs. The biggest bottleneck is often not the chips themselves, but moving data fast enough between them. If you can push 5× more data through the same physical cables, you can train bigger models faster and reduce network congestion. Why this is significant: • It shows multi-core + extended spectrum technology moving from labs into real-world commercial use • The system has already run over 35 km of existing telecom network • It could be especially useful for submarine cables and large-scale data center interconnects • China is also eyeing it for its “Eastern Data, Western Computing” project The deeper implication: We’re reaching the physical limits of how much data we can push through single-core fibers using traditional methods. By combining spatial multiplexing (multiple cores) with spectral multiplexing (more wavelength bands), engineers are finding new ways to keep scaling bandwidth without having to dig up the planet to lay new cables. This kind of breakthrough is quiet but foundational it’s the kind of infrastructure upgrade that will determine how fast AI and cloud computing can actually grow in the coming years. The future of data movement might not require more cables. It might just require smarter ones. How important do you think multi-core and multi-band fiber will be for keeping up with AI’s exploding data demands? Follow for more frontier networking, photonics, and infrastructure technology.

TheNewPhysics

20,485 views • 2 months ago

“Every child needs at least one adult who is irrationally crazy about him or her.” A couple weeks ago I shared this wonderful quote from the developmental psychologist Uri Bronfenbrenner. And it’s true. Every child needs at least one. But how lucky are those who can count a whole collection of adults who feel the same way? One of my favorite things on social media recently has been a series of videos shared to IG by gwynyth depicting not only her own interactions with her newborn son, but those of her father (aka Papa Brian). Brian’s joyful rendition of Head, Shoulders, Knees and Toes is illustrative of the loving bond they share. Check out that eye contact. Those looks of sheer delight. Frequently we discuss attachment and early relational health in terms of a parent (often a mother) and child. And while this relationship is indeed critical, the luckiest children have webs of secure, trusting relationships with multiple beloved caregivers. We can each play a role in supporting the optimal development of the young children in our lives by engaging with love, care, and our full attention. In a world where you can be anything, be like Papa Brian!
0:16

“Every child needs at least one adult who is irrationally crazy about him or her.” A couple weeks ago I shared this wonderful quote from the developmental psychologist Uri Bronfenbrenner. And it’s true. Every child needs at least one. But how lucky are those who can count a whole collection of adults who feel the same way? One of my favorite things on social media recently has been a series of videos shared to IG by gwynyth depicting not only her own interactions with her newborn son, but those of her father (aka Papa Brian). Brian’s joyful rendition of Head, Shoulders, Knees and Toes is illustrative of the loving bond they share. Check out that eye contact. Those looks of sheer delight. Frequently we discuss attachment and early relational health in terms of a parent (often a mother) and child. And while this relationship is indeed critical, the luckiest children have webs of secure, trusting relationships with multiple beloved caregivers. We can each play a role in supporting the optimal development of the young children in our lives by engaging with love, care, and our full attention. In a world where you can be anything, be like Papa Brian!

Dan Wuori

68,402 views • 10 months ago

Claude Monet painted the same stretch of cliff more than ninety times. The place is Étretat, a small fishing village on the coast of Normandy, where the chalk cliffs fall into the sea in great arches and a single spire of rock, the Aiguille, stands alone in the water. Monet had known the place since childhood. He grew up in Normandy, and these cliffs were among the first landscapes he ever saw... He returned to paint them again and again. He worked through the 1880s in front of the same rock formations, and across that time he produced more than ninety canvases of them: the cliffs at dawn, at sunset, under storm, under calm, in winter light and in the gold of a clear evening. In his letters to Alice, the woman he would later marry, he described the agony of it: the weather turning, the tide rising, the sun moving, the colour he had begun to capture vanishing before he could finish. He often worked on several canvases at once, switching between them as the conditions changed, racing each one against the hour. In a letter to his friend Frédéric Bazille he wrote: "It is beautiful here in Etretat. Every day I discover even more beautiful things. It is intoxicating me, and I want to paint it all, my head is bursting. I want to fight, scratch it off, start again, because I start to see and understand. It seems to me as if I can see nature and I can catch it all." The cliffs of Étretat had stood for millions of years and would look, to most people, the same on any given day. Monet saw that they were never the same even for two minutes. He stood on that shore and tried to hold, on canvas, something that exists only for an instant and then is gone forever. And that's exactly what those paintings really are: 90 attempts to keep a single, vanishing moment of light from disappearing. As Dylan Thomas once wrote: "Do not go gentle into that good night. Rage, rage against the dying of the light." -- -- -- If you want a deeper dive into the craft of painting, I recently wrote a piece exploring it in detail. You can read it here: And if you'd like to support my work, a paid subscription is what makes it possible:
0:17

Claude Monet painted the same stretch of cliff more than ninety times. The place is Étretat, a small fishing village on the coast of Normandy, where the chalk cliffs fall into the sea in great arches and a single spire of rock, the Aiguille, stands alone in the water. Monet had known the place since childhood. He grew up in Normandy, and these cliffs were among the first landscapes he ever saw... He returned to paint them again and again. He worked through the 1880s in front of the same rock formations, and across that time he produced more than ninety canvases of them: the cliffs at dawn, at sunset, under storm, under calm, in winter light and in the gold of a clear evening. In his letters to Alice, the woman he would later marry, he described the agony of it: the weather turning, the tide rising, the sun moving, the colour he had begun to capture vanishing before he could finish. He often worked on several canvases at once, switching between them as the conditions changed, racing each one against the hour. In a letter to his friend Frédéric Bazille he wrote: "It is beautiful here in Etretat. Every day I discover even more beautiful things. It is intoxicating me, and I want to paint it all, my head is bursting. I want to fight, scratch it off, start again, because I start to see and understand. It seems to me as if I can see nature and I can catch it all." The cliffs of Étretat had stood for millions of years and would look, to most people, the same on any given day. Monet saw that they were never the same even for two minutes. He stood on that shore and tried to hold, on canvas, something that exists only for an instant and then is gone forever. And that's exactly what those paintings really are: 90 attempts to keep a single, vanishing moment of light from disappearing. As Dylan Thomas once wrote: "Do not go gentle into that good night. Rage, rage against the dying of the light." -- -- -- If you want a deeper dive into the craft of painting, I recently wrote a piece exploring it in detail. You can read it here: And if you'd like to support my work, a paid subscription is what makes it possible:

James Lucas

57,710 views • 3 months ago

New open-source agent harness just landed! I got early access to TrueForge by TrueFoundry and have been running it locally for the past few days. The harness layer deserves as much attention as the model, and open source matters here because you can inspect the loop, run it on your own infrastructure, and swap to the latest or cheaper models. TrueForge handles the runtime work that makes an agent reliable. It drives the tool-calling loop, manages context, coordinates subagents, and executes code in a sandbox, with any model you choose. Every tool call re-sends the growing context to the model, so in practice the harness controls most of what an agent costs to run. A few things stood out from my testing and their published benchmarks. Vendor-Neutral by design. It runs OpenAI, Anthropic, and Google models alongside open-weight models like Kimi, GLM, and DeepSeek. Model routing is a setting, and you can send each task to the model that fits it. On a 14-task enterprise agent benchmark, it matched the accuracy of Claude Managed Agents running the same Opus 4.8 model at roughly 30% lower cost per run (3.8M tokens vs 10M for the same answers). Routing the same tasks to GLM-5.2 held accuracy and brought cost down by about 75%, around $3 per run instead of $12. Fully self-hosted and Open Source (MIT License). I had it running locally with one command, with sandboxed code execution working out of the box. It's time to own your agent harness. Thanks to TrueFoundry for partnering on this post.
0:16

New open-source agent harness just landed! I got early access to TrueForge by TrueFoundry and have been running it locally for the past few days. The harness layer deserves as much attention as the model, and open source matters here because you can inspect the loop, run it on your own infrastructure, and swap to the latest or cheaper models. TrueForge handles the runtime work that makes an agent reliable. It drives the tool-calling loop, manages context, coordinates subagents, and executes code in a sandbox, with any model you choose. Every tool call re-sends the growing context to the model, so in practice the harness controls most of what an agent costs to run. A few things stood out from my testing and their published benchmarks. Vendor-Neutral by design. It runs OpenAI, Anthropic, and Google models alongside open-weight models like Kimi, GLM, and DeepSeek. Model routing is a setting, and you can send each task to the model that fits it. On a 14-task enterprise agent benchmark, it matched the accuracy of Claude Managed Agents running the same Opus 4.8 model at roughly 30% lower cost per run (3.8M tokens vs 10M for the same answers). Routing the same tasks to GLM-5.2 held accuracy and brought cost down by about 75%, around $3 per run instead of $12. Fully self-hosted and Open Source (MIT License). I had it running locally with one command, with sandboxed code execution working out of the box. It's time to own your agent harness. Thanks to TrueFoundry for partnering on this post.

elvis

11,303 views • 5 days ago