AI Workflows & Agentic Architecture Directory

Nvidia released Nemotron 3.5 Lightning, a 30B-A3B MoE open model that is roughly 4x faster at output and 30% faster at agentic task completion, and open-sourced NeMo Switchyard, a routing library that picks the optimal model per request. Nvidia reports a Switchyard-routed stack cuts completion cost to about a third of running Claude Opus 4.8 alone; partners report 21% lower latency (Boomi), 58% lower cost (Ramp), and 74% lower cost at a 6% accuracy tradeoff (LangChain). Meanwhile Nemotron 4, a 1T+ parameter flagship, is in training with a ~$7B cloud-compute budget through FY2028. We map the family, model the routed unit economics, and show a routing-policy implementation.

An unreleased Anthropic model made significant — not complete — progress on the Riemann Hypothesis by testing 650 ideas across 60 parallel subagents over 31 million tokens, with findings confirmed by in-house mathematicians and formalized in Lean. The run is a blueprint for research agents that don't hallucinate proofs: an idea registry prevents duplicate exploration, and a Lean formalization gate means hypotheses only count if they compile. We diagram the fan-out pipeline, stage by stage, and price the token economics from roughly $12k to $62k depending on routing. This is progress, not a proof, and the article is precise about that.

SpaceX closed its $60 billion all-stock acquisition of Cursor on August 14, 2026, capping the coding-agent consolidation wave. With ~$4B annualized revenue, a jointly trained Colossus model shipping into Cursor and Grok Build, and AI revenue projected to lead all of SpaceX by September, here is the vendor map and the unit economics.

On August 17-18, 2026, Google transferred the Agent2Agent (A2A) protocol to the Agentic AI Foundation under the Linux Foundation, joining MCP, OpenAI's AGENTS.md, Block's goose, and agentgateway. A2A v1.0 (frozen March 12, 2026) brings signed agent cards, multitenancy, version negotiation, and multi-protocol bindings. Here is how MCP, A2A, and AGENTS.md divide the agent stack.

Cyera agreed on July 28, 2026 to acquire Oasis Security for ~$1B, bringing non-human identity (NHI) and Agentic Access Management into its data-security platform. With NHI counts in the Fortune 500 up ~500% in six months and a wave of deals — CrowdStrike-SGNL, Palo Alto-CyberArk, Cisco-Astrix — machine identity has become the agent era's control plane.

Brookfield and NextEra have proposed a $100+ billion AI-computing campus in Kentucky anchored by ~2GW of natural-gas generation and ~2.6GW of battery storage. This is utility-scale financing for AI compute — and it exposes the real ceiling on agents: firm megawatts. We break down the energy economics, the tokens-per-kWh math, and what it means for capacity planning.

Cognition, maker of the autonomous coding agent Devin, is in talks to raise at a ~$40 billion valuation in August 2026 — weeks after SpaceX bought Cursor for $60B. The AI coding arms race has moved to DevEx: time-to-first-fix, PR merge rate, senior-review handoff, and revert rate. Here is the valuation map, the DevEx scorecard, and the ROI math.

The week of August 13-16, 2026 was the biggest open-weights news week of the year: Gemini 3.7 Flash and GLM-5.3 launched, Qwen3.8-27B surfaced with no repo, Muse Spark 1.2 weights stayed unpublished, and DeepSeek moved to peak/off-peak pricing. The takeaway is discipline: announced is not shipped.

Agents belong on the warehouse floor in 2026. warehouse-mcp is a production FastMCP Python server giving AI agents six governed tools — item lookup, stock levels, order picking, restock triggers, supplier status, low-stock alerts — with inputSchema, mcpServers config, OAuth 2.0/API-key security, and retry rules for unreliable WMS backends.

DeepSeek's Aug 16, 2026 peak/off-peak rate card made time-of-day a first-class LLM cost input. This guide builds cost-router-mcp, a FastMCP Python server that routes agent calls to the cheapest eligible model — with a routing engine, cost-table JSON, provider config, token-budget caps, fallback retries, and OAuth 2.0/API-key security.

AI compute now follows power, not the other way around. This LangGraph workflow watches live energy prices and grid carbon intensity, holds training for cheap windows, gates expensive runs behind human approval, and carries a power-shutdown safety rail for interruptible-tariff campuses.

Teams run Cursor, Claude Code, and terminal agents — but nobody decides which agent runs which ticket. This LangGraph orchestrator decomposes work, routes subtasks to the cheapest capable agent via a live cost table, reviews every diff, and refuses to merge until a unified SWE-bench-style eval gate and a human sign-off.