Topic
All digests tagged LLMs
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 build a voice agent by integrating Speech-to-Text, LLMs, and Text-to-Speech. We demonstrate using AssemblyAI and LiveKit, including advanced features like tool calling and voice cloning.
Read article
Fuad Ali details the self-improvement loop for AI agents: trace failures, use Signal to detect issues, create living regression datasets, and run experiments to prove fixes. This automates reliability.
Read article
The future of AI requires mechanism design: shaping rules and affordances to build an ecosystem, not just a product. Human context and protocols are key to guiding LLMs.
Read article
Score Centering stabilizes Off-policy RL for LLMs by correcting the policy gradient loss when the training and inference policies mismatch. The method uses a Top-K approximation for efficient computation.
Read article
Building an AI code review agent requires more than a simple sandbox setup. Key fixes include model inversion, deep codebase context, autonomous testing, and cost-efficient model routing.
Read article
To ensure AI agent reliability, the speaker details a five-step regression loop. This process uses production traces as the source of truth, employing LLMs as judges to quantify subjective metrics and integrating automated gates into CI/CD.
Read article
Monzo built an in-house AI coding agent that authors 10% of PRs. Success came not from the LLM, but from building a robust, composable platform with an MCP gateway and sandboxed orchestration.
Read article
The panel explores why OAuth is insufficient for autonomous AI agents. The solution lies in 'intent-based authorization,' which uses mission context and architectural separation (brain vs. hands) to manage risk.
Read article
AI agents are cutting 3-day portfolio builds to minutes in structured credit. Learn how Apollo is building AI-native platforms to automate complex, fragmented financial data reconciliation.
Read article
AI makes building software easier than ever, but this convergence creates a 'median' product that is free for everyone. To succeed, focus on novelty and distribution, not just the model.
Read article
DDC is a methodology that uses AI agents to fail on real production incidents, systematically identifying and curating the minimum necessary knowledge context. This approach treats incident data as knowledge tests, moving beyond static runbooks to a 'living context.'
Read article
The future of AI products lies in 'always-on' agents that orchestrate themselves. We explore the shift from completions to multi-agent systems, demonstrating how agents can run complex, end-to-end businesses.
Read article
Modular addresses the AI software mess by introducing Mojo, a Pythonic systems language, and MAX, a unified inference framework. This stack enables portable, high-performance deployment across mixed, multi-node hardware.
Read article
Learn how AI agents use real-time web data from sources like Amazon and Reddit to build product intelligence recommendations. We cover the shift from brittle scraping pipelines to scalable, resilient data collection using Bright Data's MCP.
Read article
AI agents pose a 'lethal trifecta' threat: private data access, untrusted content exposure, and external communication. Secure adoption requires three guardrails: CDEs, a Model Proxy, and an Agent Firewall.
Read article
Running AI agents in production requires separating the 'brain' (LLM planning) from the 'hands' (deterministic harness). This approach treats agent harnesses as late-bound sagas, ensuring reliability and durability for complex, long-running workflows.
Read article
Learn how to build an intelligent robot arm from scratch, moving from basic servos and microcontrollers to advanced AI integration using Physical AI principles.
Read article
Building agent quality platforms is a complex systems challenge. Learn why simple spreadsheets fail and how to build the continuous improvement loop using advanced tracing and observability.
Read article
CoreWeave Forge demonstrates the full AI loop: Observe, Curate, Improve, and Evaluate. Engineers can systematically improve AI agents using Agent Lens, Serverless RL, and model lineage tracking.
Read article
OpenRouter usage shows token spend is up 80x year-over-year. The key to AI success is 'tokconomics': optimizing workflows using specialized models like Jev alongside LLMs to maximize value per token.
Read article