How I use Cursor, Copilot, and Claude on real client work: mapping an unfamiliar legacy codebase, moving a decade-old build stack to Vite, and getting tests that cover failure modes instead of just the happy path.
WordPress now has a repository for AI agent skills, starting with wp-playground. It uses the Playground CLI so an agent can mount a plugin or theme, wait for WordPress to be ready, log in, and check its own work in seconds.
Data science still gets treated as an academic exercise. Engineering constraints, caching, timeouts and safe fallbacks are what turn a model into something you can actually run on a client site.
Raw JSON in an embedding model spends your token budget on braces and quotes. Flattening product data into plain sentences gives the model the kind of text it was trained to read.
A notebook demo is not production. This is what sits between the two: an API contract, a container, infrastructure that suits the workload, and monitoring that catches drift before the predictions quietly go bad.
Metadata filters only take music discovery so far. How CNNs read audio as Mel-spectrograms, how contrastive learning with InfoNCE loss shapes the embeddings, and why the inference belongs outside WordPress.
Fastest-route logic says nothing about what a walk is actually like. How the StreetSense project models urban walking risk with Uber’s H3 indexing, sine and cosine time encoding, and XGBoost under a Tweedie objective.
Multi-GPU training usually stalls on communication, not compute. What nvidia-smi topo and Nsight show you about NCCL overhead, and the DDP settings I change: static graph, bucket views, PowerSGD compression and smaller buckets.
Dongxiang has almost no parallel data, so we fine-tuned Meta’s NLLB-200 instead of prompting a general model. Cleaning the bilingual corpus, checking subword fertility, registering a new language tag, and why Adafactor beat AdamW on one GPU.
Notes from the AI contributor meeting: AI Experiments 0.2.0 is out, the Core Post Management abilities PR is under review, PHP AI Client v0.4.0 shipped, and the MCP Adapter stays a canonical plugin. February 19 is the Beta 1 cutoff.