AIE Miami Day 2 ft. Cerebras, OpenCode, Cursor, Arize AI, and more!
summary
Latency debt is measurable: wafer‑scale inference demos claim 1200 tokens per second code generation and advertise Codex Spark as a step‑change for fast, pair‑programming style coding. Practical claims and builds matter: offline latent diffusion runs on mobile NPUs with ~600 ms inference; ambient sensors map to latents; Cursor ships Cursor 3.0 promising IDE simplification and visual workflows; Arize publishes head‑to‑head evals comparing MCP vs CLI for agent tool integration; multiple teams report hundreds to thousands of production agent users.
AI-written summary · about
chapters (21)
- 0:00Intro
- 1:01Opening remarks by Ethel Zhang and Iman Haji
- 4:45David House (G2i): agentic coding case studies for mindset shifts and expert amplification
- 26:34Sarah Chieng (Cerebras): "latency debt", Wafer Scale inference and 1200 tokens per second code generation
- 44:29Lech Kalinowski (CallStack): offline latent diffusion on mobile NPUs with ambient light prompts
- 1:00:00Break: Morning coffee
- 1:51:37Tejas Bhakta (Morph LLM): "Software 3.5", specialized models beating frontier compute constraints
- 2:04:36Rick Blalock (Agentuity): coding agents as the universal software primitive for business workflows
- 2:27:02Nyah Macklin (Neo4j): context graphs for auditable, explainable AI decision traces
- 2:58:24Break: Lunch
- 4:25:48Gabe Greenberg (G2i): NVIDIA DGX Spark giveaway for a frontier lab intro
- 4:29:07Lena Hall (Akamai): Reachy Mini, embodied AI and real time multimodal robot behavior
- 4:51:47Dave Kiss (Mux): designing SaaS infrastructure and auth for agents as first class users
- 5:17:08Alvin Pane (OutRival): agentic memory replacing RAG pipelines for long term memory SOTA
- 5:35:07Erik Thorelli (CodeRabbit): skills, error loops, and continuous learning for agent systems
- 5:56:02Break: Afternoon
- 6:38:50Hassan El Mghari (Together AI): open source models powering AI apps with millions of users
- 6:57:19Stefan Avram (OpenCode): multi-model AI coding tools and open source model flexibility
- 7:08:44Laurie Voss (Arize AI): evals comparing MCP and CLI tool patterns for agents
- 7:37:12David Gomes (Cursor): Cursor 3.0, IDE simplification, and AI driven visual workflows
- 7:54:13Closing remarks by Ethel Zhang and Iman Haji
related talks
20mAI editors – how they work, and how to master them
37mSoftware Engineering + AI = ? | The Future of Development
07h 47mAI Agents Conference 2026 - Day 5
26mFrom autocomplete to agents: AI coding assistance state of play
27mSoftware engineering with GenAI
51mLive Panel Discussion: Less Code? More Vibes? Web Dev in the AI Era
29mGenAI in Backstage - Turbocharge Your Developer Productivity!
35mUsing coding agents with open models