[ 68429340.000 ] posts.x: Full talk on distilling successful and failed agent runs into reusable skills:  |   [ 68429340.000 ] posts.x: A Chromium fork running inside AWS Lambda succeeded 36% of the time, not exactly production ready. @rafalwilinski, founding engineer at Runlayer…  |   [ 68422380.000 ] posts.x: @andimarafioti Let’s hope not!  |   [ 68402460.000 ] posts.x: Full talk on Gemini's Interactions API and why agents get their own sandboxed environment:  |   [ 68402460.000 ] posts.x: Giving an agent a function result by stuffing it into a "user" turn always felt backwards, since the result comes from the environment, not the user…  |   [ 68395680.000 ] posts.x: @levelsio  |   [ 68395620.000 ] posts.x: @andimarafioti Marital discord as a result of Anthropic A/B testing! 😜  |   [ 68395560.000 ] posts.x: @thekitze What are you printing?  |   [ 68395380.000 ] posts.x: @ThePrimeagen 💀  |   [ 68395260.000 ] posts.x: @0x0SojalSec @CoffeeVectors Open models FTW  |   [ 68333520.000 ] posts.x: Full workshop on what happens inside an inference engine between prefill and decode, from scheduler bottlenecks to speculative decoding:  |   [ 68333520.000 ] posts.x: Training is a cost center. Inference is a revenue center. @charles_irl makes that case the spine of "What Is an Inference Engine, Anyway?", and…  |   [ 68322600.000 ] posts.x: Full talk on scaling databases for AI-driven traffic, covering isolation, sharding, back pressure, and letting agents draft and deploy schema changes:  |   [ 68322600.000 ] posts.x: A couple of months back GitHub's traffic jumped two to four times over what was already massive load, something they pinned largely on AI agents…  |  

[ 68429340.000 ] posts.x: Full talk on distilling successful and failed agent runs into reusable skills:  |   [ 68429340.000 ] posts.x: A Chromium fork running inside AWS Lambda succeeded 36% of the time, not exactly production ready. @rafalwilinski, founding engineer at Runlayer…  |   [ 68422380.000 ] posts.x: @andimarafioti Let’s hope not!  |   [ 68402460.000 ] posts.x: Full talk on Gemini's Interactions API and why agents get their own sandboxed environment:  |   [ 68402460.000 ] posts.x: Giving an agent a function result by stuffing it into a "user" turn always felt backwards, since the result comes from the environment, not the user…  |   [ 68395680.000 ] posts.x: @levelsio  |   [ 68395620.000 ] posts.x: @andimarafioti Marital discord as a result of Anthropic A/B testing! 😜  |   [ 68395560.000 ] posts.x: @thekitze What are you printing?  |   [ 68395380.000 ] posts.x: @ThePrimeagen 💀  |   [ 68395260.000 ] posts.x: @0x0SojalSec @CoffeeVectors Open models FTW  |   [ 68333520.000 ] posts.x: Full workshop on what happens inside an inference engine between prefill and decode, from scheduler bottlenecks to speculative decoding:  |   [ 68333520.000 ] posts.x: Training is a cost center. Inference is a revenue center. @charles_irl makes that case the spine of "What Is an Inference Engine, Anyway?", and…  |   [ 68322600.000 ] posts.x: Full talk on scaling databases for AI-driven traffic, covering isolation, sharding, back pressure, and letting agents draft and deploy schema changes:  |   [ 68322600.000 ] posts.x: A couple of months back GitHub's traffic jumped two to four times over what was already massive load, something they pinned largely on AI agents…  |