[ 68869740.000 ] posts.x: Full talk on why agents need a persistent graph of what data actually means, not just access to the data, plus a worked example of an agent reaching…  |   [ 68869740.000 ] posts.x: Give an agent access to what a table's rows actually mean, and task success on complex data changes jumps almost 35 points over the same model…  |   [ 68847720.000 ] posts.x: Full talk on the document extraction pipeline behind JMail, from redaction cleanup to schema design:  |   [ 68847720.000 ] posts.x: Over three million pages of redacted, scanned, and handwritten government documents sat between an internet joke and a real extraction problem…  |   [ 68833800.000 ] posts.x: Full talk on building evaluators for a sample support agent and using a coding agent to mine production traces for failures:  |   [ 68833800.000 ] posts.x: An agent's dashboards can look perfectly green while its actual answers are wrong, and that gap between system health and response quality is what…  |   [ 68825580.000 ] posts.x: Full talk on owning versus renting GPU inference, the KV cache math behind it, and where quantization and speculative decoding actually help:  |   [ 68825580.000 ] posts.x: Speculative decoding is supposed to make inference faster. @aiDotEngineer posted "Agents That Own Their Inference" to YouTube, where @labeveryday, a…  |   [ 68784000.000 ] posts.x: Full talk on hardening an agent-driven platform, including the one attack (a fork bomb) no model would ever run:  |   [ 68784000.000 ] posts.x: Rename a malicious container image from "Joker" to "promo-mascot" and the exact same model that refused to deploy it suddenly ships it without a…  |   [ 68771580.000 ] posts.x: Full workshop on wiring Qodo's rules, skills, and local gates into a payments API refund task:  |   [ 68771580.000 ] posts.x: A refund endpoint that keeps its idempotency state in a process local dictionary isn't durable storage. That's the bug @nnennahacks, Principal…  |   [ 68764080.000 ] posts.x: Full talk on MiniMax's sparse attention design and why multimodality got trained in from the start instead of added after:  |   [ 68764080.000 ] posts.x: Most sparse attention designs look clean on paper and then run into real GPU memory access patterns. @olive_jy_song, RL Lead at MiniMax, works…  |  

[ 68869740.000 ] posts.x: Full talk on why agents need a persistent graph of what data actually means, not just access to the data, plus a worked example of an agent reaching…  |   [ 68869740.000 ] posts.x: Give an agent access to what a table's rows actually mean, and task success on complex data changes jumps almost 35 points over the same model…  |   [ 68847720.000 ] posts.x: Full talk on the document extraction pipeline behind JMail, from redaction cleanup to schema design:  |   [ 68847720.000 ] posts.x: Over three million pages of redacted, scanned, and handwritten government documents sat between an internet joke and a real extraction problem…  |   [ 68833800.000 ] posts.x: Full talk on building evaluators for a sample support agent and using a coding agent to mine production traces for failures:  |   [ 68833800.000 ] posts.x: An agent's dashboards can look perfectly green while its actual answers are wrong, and that gap between system health and response quality is what…  |   [ 68825580.000 ] posts.x: Full talk on owning versus renting GPU inference, the KV cache math behind it, and where quantization and speculative decoding actually help:  |   [ 68825580.000 ] posts.x: Speculative decoding is supposed to make inference faster. @aiDotEngineer posted "Agents That Own Their Inference" to YouTube, where @labeveryday, a…  |   [ 68784000.000 ] posts.x: Full talk on hardening an agent-driven platform, including the one attack (a fork bomb) no model would ever run:  |   [ 68784000.000 ] posts.x: Rename a malicious container image from "Joker" to "promo-mascot" and the exact same model that refused to deploy it suddenly ships it without a…  |   [ 68771580.000 ] posts.x: Full workshop on wiring Qodo's rules, skills, and local gates into a payments API refund task:  |   [ 68771580.000 ] posts.x: A refund endpoint that keeps its idempotency state in a process local dictionary isn't durable storage. That's the bug @nnennahacks, Principal…  |   [ 68764080.000 ] posts.x: Full talk on MiniMax's sparse attention design and why multimodality got trained in from the start instead of added after:  |   [ 68764080.000 ] posts.x: Most sparse attention designs look clean on paper and then run into real GPU memory access patterns. @olive_jy_song, RL Lead at MiniMax, works…  |