AI coding prompts linked to higher costs in benchmark study
A preprint examines prompt detail, thinking effort and low-cost probes as predictors of agentic coding costs for unseen tasks in a benchmark.
Topics: Coding Agents
Entities: Coding Agents
A preprint examines prompt detail, thinking effort and low-cost probes as predictors of agentic coding costs for unseen tasks in a benchmark.
Topics: Coding Agents
Entities: Coding Agents
How Python does AI. Agents, realtime voice, image generation, embeddings. Every model, every interface, typed end to end. - Release v2.37.0 (2026-08-31) · pydantic/pydantic-ai
A preprint reports an ImageNet system that sent less compressed data than WebP and BPG at a stated image-classification accuracy.
A preprint presents BVR Sim, an open air-combat simulator with faster C++ throughput and limited policy transfer across different aircraft models.
A preprint reports a learned simulator that approximates stochastic reaction dynamics with fewer steps than SSA in tested numerical networks.
A new preprint models whether Punctum could be a primordial black hole, using its millimeter properties to estimate dark-matter annihilation parameters.
A mathematical preprint links p-adic reduction of abeloid varieties with monodromy, Galois representations and conjugate uniformization.
New from Graham Dumpleton (of wrapt, mod_wsgi, and New Relic's Python agent fame), who describes Wrapture as taking the monkeypatching ideas from wrapt and extending them to apply to testing …
Topics: Developer Automation
Entities: Developer Automation
A preprint tests GIFT, a GPU system designed to isolate user data in LLM serving, and reports overhead in selected model and workload benchmarks.
Keenable AI open-sources NEEDLE, a live benchmark that rebuilds search queries hourly and scores 15 APIs for agents.
A preprint reports higher combined function-and-security scores for a four-agent code system on CWEval and BaxBench benchmark tests.
A theoretical preprint examines temperature-linked spin dynamics and a proposed non-Abelian gauge potential in a ferromagnet, without experimental testing.
Topics: Developer Automation
Entities: Developer Automation