Gutenberg 22.6: icons, revisions, and real-time collaboration
Gutenberg 22.6 adds a server-side Icon block, visual revision diffing, and real-time collaboration via Yjs. It’s a stability-focused release ahead of WordPress 7.0.
Tag: APIs
Gemini object detection instead of training a custom model
Gemini’s open-vocabulary detection removes the label-thousands-of-images step from a vision pipeline. Notes on structured JSON output with Pydantic, editing detected regions with Nano Banana, and the rate limits that show up in production.
How to define an internal credit risk model modeling scope
Ahmad Wael walks through how to define an internal credit risk model modeling scope: building the (ID x Year) dataset, avoiding temporal overlaps, and meeting ECB standards for PD, EAD, and LGD.
Gradient accumulation before you buy more GPUs
Getting a larger effective batch size out of the GPUs you already have: gradient accumulation in a PyTorch loop, DDP with no_sync(), why the loss has to be scaled, and the bucket size and data loading settings that decide whether scaling stays linear.
How WordPress 7.0 changes the iframed editor for custom blocks
WordPress 7.0 changes how the iframed editor decides which posts get isolated CSS, checking blocks actually used in a post rather than every block registered on the site. Legacy blocks still work, but the push toward Block API version 3 is real.
Causal reasoning models behind Nvidia’s AlpamayoR1
A read of Nvidia’s AlpamayoR1: the Cosmos-Reason backbone, the joint action-reasoning token space that keeps an explanation tied to the steering, GRPO post-training, and why benchmarks run on Nvidia’s own datasets are hard to trust.
WordPress core performance: real-time editing vs object caching
The WordPress 7.0 collaboration prototype polls up to four times a second, and every poll wipes the object cache for the post it touches. Here is what Ticket #64696 found and why core is getting its first new table since 2015.
PySpark for Pandas users: getting past the RAM wall
Pandas keeps the whole dataset in RAM and computes as you type. PySpark builds a plan first and runs it across a cluster. Here is what that changes for loading data, window functions, and the shuffle settings nobody tunes.
Streak system design: timezones, race conditions and grace
A streak is not a counter you increment. Notes on the psychology that makes streaks work, the timezone and race condition bugs that break them, and why grace mechanisms belong in the design from the start.
What the CSS :near pseudo-class would replace in your JS
Thomas Walichiewicz proposed a :near() selector so CSS can react when a pointer gets close to an element. It would replace the usual pointermove distance math, though the CSS Working Group is still weighing the fingerprinting risk.