The AI landscape is rapidly evolving, marked by IBM's B300 cluster partnership, Meta's local Muse Glimmer model, and OpenAI's advanced Astra capabilities. We explore the shift to edge computing and industrial-scale infrastructure.
Read article
See how Gemini 3.7 Flash is used to build a playable animated sprite game in Google Antigravity from a single prompt, showcasing rapid prototyping capabilities.
Read article
Gemini 3.7 Flash enables agent-first workflows, allowing users to build complex games and prototypes in Google Antigravity from single prompts.
Read article
The lecture details model-based RL, emphasizing that when using learned dynamics for planning, quantifying uncertainty (epistemic) is crucial. Techniques like Bayesian statistics and ensembles allow computing expected rewards by averaging over a posterior distribution of possible models.
Read article
Dive into model-free Reinforcement Learning! We explore value-based methods like SARSA and Q-learning, contrasting on-policy vs. off-policy approaches. Learn how Deep Q Networks (DQN) stabilize training using Experience Replay and fixed target networks.
Read article
MPC solves an open-loop optimal control problem over a finite horizon at each time step. By using only the first computed input and recomputing everything from scratch (receding horizon), it achieves closed-loop performance while maintaining computational tractability.
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
Great models often fail in production due to poor scoping, monitoring gaps, or changing real-world data. Success requires aligning model value with business need and planning for drift.
Read article
This talk details building agentic systems using a modified Doom-like game world as a controlled testbed. The architecture emphasizes observability, determinism, and the use of semantic mirrors for complex AI simulation.
Read article
The talk explores how LLMs impact software expertise. While adoption is high, researchers warn that over-reliance can degrade skills, impair critical thinking, and lead to 'cognitive surrender,' posing a risk to junior engineers.
Read article
Learn how AI and Data Science combine into a single pipeline. We break down complex concepts like RAG, embeddings, and data cleansing using periodic tables to show how robust enterprise AI applications are built.
Read article
Centralized LLM Gateways are critical for governing autonomous agents, managing cost, ensuring compliance (GDPR, HIPAA), and providing reliability through fallbacks and rate limits across the entire Agent Development Lifecycle.
Read article
This lecture extends optimal control to stochastic settings using Markov Decision Processes (MDPs). The core methodology involves defining state transitions and costs that incorporate random disturbances ($w_k$).
Read article
LQR is generalized from state regulation to nonlinear tracking by defining deviation variables ($\delta x, \delta u$). This allows the problem to be reformulated as an auxiliary LQR problem.
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
This analysis explores how ChatGPT and Claude implement memory via running profiles. The speaker argues that memory is a function of compute and cannot be outsourced.
Read article
Learn how to build an autonomous social media agent using LangChain's Managed Deep Agents. We cover tool creation, scheduling, memory management, and full deployment to Slack.
Read article
For modern AI agents, evaluation must shift from prompt playgrounds to robust code tests. Focus on 'raising the floor'—mitigating worst-case failures by tracking issue start dates and user impact.
Read article
Learn how Applied Compute brings continual learning to enterprises using a 2x2 distillation spectrum. Discover how online hinting and per-step guidance can boost agent performance from 15% to 80% without golden answers.
Read article
This session details the Model Cascade architecture, a technique to drastically reduce LLM classification costs (by over 90%) while maintaining high accuracy (e.g., 95%+). Instead of running an expensive 'Oracle' model on every record, the cascade uses a cheap proxy model for initial predictions and only escalates to the Oracle when the proxy's confidence score is low. The optimal decision threshold ($\tau$) is found offline by simulating performance across samples.
Read article