Apple releases SCLATE, a training and evaluation harness for continual-learning agents
09/30/2026 — 10/04, 02:11·1 sources·1 reports
Apple introduced SCLATE, an execution harness for training and evaluating continual-learning agents. Such agents combine a model, a harness, and memory across multiple sessions. While existing benchmarks and training frameworks only schedule their own events, SCLATE interleaves task and session lifecycle events, cron jobs, and memory consolidation on the agent side.
Latest turnSCLATE is an execution substrate for training and evaluating continual-learning agents, which combine models, harnesses and memory over long multi-session horizons. Existing benchmarks and training frameworks only schedule their own events, forcing every benchmark-agent pair to build a custom scheduling loop; SCLATE interleaves tasks with agent-side events such as session stops and starts, crons and memory consolidation.
Reports on this story headlines open the original
SCLATE is an execution substrate for training and evaluating continual-learning agents, which combine models, harnesses and memory over long multi-session horizons. Existing benchmarks and training frameworks only schedule their own events, forcing every benchmark-agent pair to build a custom scheduling loop; SCLATE interleaves tasks with agent-side events such as session stops and starts, crons and memory consolidation.
Apple Machine LearningFirst-partyAI score 60
Other stories people are talking about
- 519RisingApple tightens macOS Full Disk Access over AI agent risks9 sources
- 368NVIDIA launches 64GB DGX Spark desktop AI computer at $4,9998 sources
- 353RisingMeta open-sources Muse Gadgets firmware and SDK6 sources
- 271Claude Opus 5.5 and GPT-6 Sol: comparing cost per task4 sources
- 248Anthropic Reportedly Targets Nov 9 IPO Listing Before Thanksgiving5 sources
- 234Aleph Alpha releases Kolibri-1, a 78B MoE model with 1M context3 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion