Abundant Blog https://abundant.ai/blog/ Thoughts from the Abundant team. en-us Wed, 27 May 2026 07:00:00 +0000 RL environment creation is becoming continuous QA https://abundant.ai/blog/env-creation-loop/ https://abundant.ai/blog/env-creation-loop/ Wed, 27 May 2026 00:00:00 -0700 Rishi Desai Why automating RL environment creation means building the loop around task generation. Designing eval harnesses that prevent reward hacking https://abundant.ai/blog/task-design/ https://abundant.ai/blog/task-design/ Wed, 13 May 2026 00:00:00 -0700 Rishi Desai Trust boundaries for coding agents: verifiers, artifacts, and network access. Frontier Models Caught Cheating https://abundant.ai/blog/frontier-models-caught-cheating/ https://abundant.ai/blog/frontier-models-caught-cheating/ Thu, 16 Apr 2026 00:00:00 -0700 Joan Cabezas + Abundant Research Team Frontier coding agents caught cheating on long-horizon tasks: a leaderboard, a taxonomy of reward hacks, and what it means for benchmarks. Hillclimbing to an abundant future https://abundant.ai/blog/hillclimbing-to-an-abundant-future/ https://abundant.ai/blog/hillclimbing-to-an-abundant-future/ Sun, 05 Apr 2026 00:00:00 -0700 Jesse Hu Hillclimbing — the practice of making a number go up for a capability that resists clean definition — is the core bottleneck on the path to AGI.