Topic
All digests tagged LLM Optimization
Suggested route
Move from earlier context through the available material to the newest signals. This route is a reading aid, not a difficulty ranking.
Learn how to 'actually tokenmaxx' LLMs. This guide details a systematic workflow using DSPy and shadow evaluation to cut prompt bloat by 51% while maintaining quality and reducing compute costs.
Read article
The future of enterprise AI lies in long-running agents that automate document workflows. We explore the advanced document context layer, covering specialized OCR, agent harnesses, and optimizing models for cost and accuracy.
Read article
This devlog details building a self-compacting Pi Agent to solve context window limits in autonomous agents. By giving the agent a dedicated tool and three compaction thresholds (Notice, Warning, Force), engineers can achieve reliable, out-of-loop agentic coding.
Read article
The talk details the 'big three' of inference engineering: Quantization, KV compaction, and Speculation. Learn about DFlash, TurboQuant, and how training is merging with inference optimization.
Read article
RL is revolutionizing search by creating specialized sub-agents. These agents can adaptively search, leading to results that are 20 times faster and 100 times cheaper than using frontier models.
Read article
CoreWeave ARIA is an AI research agent built into Weights & Biases that accelerates the self-improving AI loop. It automates auto-research, analyzes training runs, and optimizes agent performance.
Read article
NeMo Switchyard is an open-source model routing library for AI agents that automatically routes each query to the best available model. It solves the 'one model' problem by optimizing selection across local/cloud and open/closed models.
Read article
Discover how Gemini 3.7 Flash powers scalable multi-agent systems for enterprise use. Learn about optimizing costs (up to 10x cheaper) and implementing low-latency intent classification.
Read article
Mastering the System Prompt is key to controlling AI agents. Learn how to use positive/negative patterns, reference points (D1/R6), and hard operational boundaries to turn verbose LLMs into precise senior engineers.
Read article
Unify achieved a 90-95% cost optimization by treating subagents as simple function calls. Key strategies include optimizing prompt cache hits, using specialized REPLs like Monty for durable cloud execution, and implementing Pandas-like wrappers over databases.
Read article
Improving agents is a data mining problem. The key to continuous improvement is collecting comprehensive agent traces, which serve as the substrate for observability and continual learning.
Read article
The industry is shifting focus from training bigger models to optimizing cost and speed. Key techniques include the 'sidekick' agent architecture, which uses parallel processing of multiple LLMs for better price performance.
Read article
Learn how to move from token maxxing to value maxxing with GPT-5.6. Optimize your AI agents by leveraging programmatic tool calling, prompt caching, and model selection (Sol, Terra, Luna) for maximum cost efficiency.
Read article