The Brief

New briefs daily around 7 AM Eastern

Save tokens by having smaller models handle search

Search consumed more than half the tokens across 1,200 coding-agent traces. Relace's dedicated search model keeps accuracy near a frontier model while cutting latency 4x, leaving the expensive model to write code.

• 01Read

  Compacting on cache miss cut one agent trace 55 percent

  For long-running coding agents, input tokens now dominate spend. Compacting context at the cache miss turns cache-read cost from quadratic to linear, measured at 55 percent off one real trace. The constraint is that compaction only pays if it runs fast enough to stay invisible to the agent loop.

  Relace→
• 02Read

  Mercury 2 decodes by parallel refinement, not left to right

  Inception's reasoning model reaches four-figure tokens per second on standard NVIDIA GPUs by refining a draft in parallel rather than committing token by token. The pitch is that test-time compute stops trading against latency, which is what puts a reasoning model inside a 500ms voice budget that GPT-4.1 currently holds by default.

  Introducing Mercury 2→Mercury 2 The First Reasoning Model Fast Enough To Pick Up The Phone→
• 03Read

  Mercury Edit 2 uses KTO to make edits 27 percent more selective

  Next-edit models fail by suggesting too much, not by suggesting wrongly. Aligning on logged accept and reject signals with KTO made edits 48 percent more likely to be accepted alongside the selectivity gain, which treats suggestion fatigue as an alignment target rather than a UI problem.

  Inception→
• 04Read

  Reka Speech transcribes 8 to 35x faster than WhisperX

  An 850M model produces timestamped transcription and translation on a single H100, offloading query and key embeddings and recovering alignment after generation instead of running a separate forced-alignment pass. It is a bet that timestamped transcription at scale is throughput-bound rather than accuracy-bound.

  Reka AI→
• 05Discuss

  Reka argues agent scaffolds substitute for reasoning models lack

  Reka's position is that template orchestration, where an LLM is called repeatedly to choose the next step, carries only shallow reasoning, and that decomposition, self-verification and backtracking should be trained into the model instead of supervised at inference time. That is a direct challenge to how most agent harnesses shipping today are built.

  Reka AI→
• 06Read

  SemiAnalysis weighs OpenAI's Jalapeno chip against CUDA lock-in

  The breakdown covers performance per megawatt against Blackwell, where the published benchmark comparisons stop holding, Samsung HBM4 supply, and how far AI-assisted design carried the project. It is the same cost pressure driving work off frontier models, read one layer down in the stack.

  SemiAnalysis→