4 stories about DeepSeek.
DeepSeek released a design article and a prototype repo for a long-lived router agent that hands work to isolated workers triggered by GitHub events.
Why it matters: Companies building multi-agent coding workflows get a concrete, open reference for routing work across parallel workers, but the stated gaps in memory and security mean it is not yet production ready.
ISTA DAS Lab's experimental build keeps half the experts per layer and uses low-bit quantization, reporting 98.7% and 94.8% of base benchmark scores.
Why it matters: Pruned and quantized large coding models could fit on smaller hardware, but the reproducibility questions mean buyers should verify results before relying on them.
DeepSeek has adapted its internal Super-agent for public research use, letting it watch a GitHub repo and start agent sessions that can return pull requests.
Why it matters: Engineering teams get another open option for event-driven coding agents, which could reduce the manual work of routine repository tasks.
A 50-task benchmark of games and web tasks shows top models pass only 36% of tasks under real-time latency, versus up to about 80% with a paused clock.
Why it matters: The gap between paused and real-time results suggests current agents are unreliable wherever response speed matters, such as buying tickets or bidding in auctions. Teams should not assume that strong offline benchmark scores carry over to live, latency-sensitive workflows.