iFeeling Daily
Daily curated AI insights you can't miss.
Building Six Small Models with ML Intern for About USD 103
ML Intern in HuggingChat built six small models from one-line ideas, with strict budgets, baselines, and smoke tests, for about USD 103 total in compute.
OpenAI Disrupts Russian and Iranian False-Front Influence Operations
OpenAI disrupted two false-front influence operations, including its first Category 5 case, and detailed AI's role in deceptive personas and media seeding.
Understanding Task Costs in Claude Code on Opus 5.5
Opus 5.5 cuts token prices, but how much you actually spend depends on turns, cache reads, output, and the model you choose for your real tasks.
How Effort Levels Change Claude Code’s Output
Effort in Claude Code trades speed for deeper verification: low effort iterates fast, while high and max effort catch edge cases but cost far more time and tokens.
Automating eval design and hillclimbing with Claude
The claude-api skill adds two workflows: build-eval to design evaluations and hillclimb to improve performance against them with holdout checks.
Citrix NetScaler Zero-Day Exploitation: CVE-2026-88772 and Campaign Details
Mandiant and GTIG report active exploitation of Citrix NetScaler zero-days CVE-2026-88772 and CVE-2026-88771, detailing attack chains, persistence, and hunting guidance.
GPU/TPU Utilization with Multi-Cluster GKE Inference Gateway
Google's multi-cluster GKE Inference Gateway pools globally scattered GPU/TPU capacity into one fleet, achieving near-linear scale and 99.5% throughput with less than 1% overhead.
Async GRPO with LoRA across HF Jobs: A Bucket, a Proxy, No NCCL
AsyncGRPOTrainer now syncs only a rank-1 LoRA adapter to vLLM via a shared Storage Bucket, so training and inference run in separate HF Jobs with no NCCL.
How OpenAI Scaled Habitat Storage to 1 Billion Weekly Users
OpenAI's Habitat online storage now handles over 70 million requests per second and 500+ petabytes, supporting 1 billion weekly users.