--- title: "Aldena blog: notes on running a team of AI agents" description: "notes from building aldena: how a team of ai agents with a project manager and a reviewer ships features, and how that compares to a single coding agent." url: "https://aldena.ai/blog" --- # notes on running a team of AI agents what I learned building Aldena: roles, handoffs, reviews, and where a single coding agent runs out of road. - [llama vs gpt vs claude: pick a model per role, not a winner](https://aldena.ai/blog/llama-vs-gpt-vs-claude): 2026-08-26. an open vs closed model comparison for choosing what runs your ai agents: cost, latency, eval deltas, and why per-role assignment beats one champion. - [claude sonnet 5 benchmarks: what the scores mean for agents](https://aldena.ai/blog/claude-sonnet-5-benchmarks): 2026-08-25. sonnet 5 beats opus 4.8 on terminal work and edges it on knowledge work at 40 percent less. i map each launch benchmark to the agent job it actually predicts. - [claude opus 4.5 benchmarks: which numbers predict quality](https://aldena.ai/blog/claude-opus-4-5-benchmarks): 2026-08-24. opus 4.5's reported scores, read by someone who runs agent teams on these models: which benchmarks predict delegated work, and which ones are noise. - [what it actually takes to self-host an always-on ai agent](https://aldena.ai/blog/self-hosted-ai-agent): 2026-08-23. the real bill of materials behind a self-hosted ai agent: gateway, sandbox, memory, browser, auth, and cost. and what changes when that job is someone else's. - [Your first AI employee: what changes in week one](https://aldena.ai/blog/first-ai-employee): 2026-08-22. Hiring an AI employee is a shift from tools you operate to teammates that deliver. What to delegate first, and why a small team beats a lone hire. - [An AI employee platform: hire roles, not build workflows](https://aldena.ai/blog/ai-employee-platform): 2026-08-21. What an AI employee platform does for a small team, where human review has to sit, and why hiring a role beats mapping a workflow. - [AI-assisted development: what changed on my team](https://aldena.ai/blog/ai-assisted-development): 2026-08-20. An honest report from building Aldena with its own agent teams: the roles, the guardrails, the review gate that stays human, and where the money goes. - [AI coding agents: direct and verify, don't ask and paste](https://aldena.ai/blog/ai-coding-agents-guide): 2026-08-19. A practical guide to getting real work out of AI coding agents. Plan before you delegate, scope small, review with suspicion, and trust tests over confidence. - [Multi-agent platforms: what to demand before you buy one](https://aldena.ai/blog/multi-agent-platforms): 2026-08-18. A buyer's guide to multi-agent platforms. How a platform differs from a framework, what to test in orchestration and permissions, and where guardrails go. - [Enterprise AI agents pay off when they become teams](https://aldena.ai/blog/enterprise-ai-agents): 2026-08-17. The business case for enterprise AI agents is teams, not smarter assistants. Where the value shows up, what governance has to exist, and how I would start. - [ai employees: what the term means and what it does not](https://aldena.ai/blog/ai-employees): 2026-08-16. an ai employee is an agent with a role, permissions, memory, and a review gate. I draw the line against chatbots, RPA and copilots, and show a real deployment. - [An AI workforce platform is a control plane, not a chatbot](https://aldena.ai/blog/ai-workforce-platform): 2026-08-15. What an AI workforce platform does, how hiring an AI worker differs from building one, and how Aldena models a workforce with rooms, hierarchy, and approvals. - [AI agent frameworks: what they give you, and what they don't](https://aldena.ai/blog/ai-agent-frameworks): 2026-08-14. What an AI agent framework actually provides, how crewAI, LangGraph, AutoGen, and the provider SDKs differ, and the infrastructure none of them ship. - [what is a multi-agent system? definition and components](https://aldena.ai/blog/what-is-a-multi-agent-system): 2026-08-13. a multi-agent system is several autonomous agents coordinating in a shared environment. the definition, the components, the use cases, and the failure modes. - [Multi-Agent Systems: How They Work and Who Needs One (2026)](https://aldena.ai/blog/multi-agent-systems): 2026-08-12. How multi-agent systems decompose work, coordinate, and fail in production, plus a plain test for when a single agent is the cheaper, better call. - [AI agent orchestration in production: what actually breaks](https://aldena.ai/blog/ai-agent-orchestration): 2026-08-11. How I orchestrate AI agent teams in Aldena: why hierarchy beats a flat swarm, how work travels, what shared memory and approval gates do, and where tokens go. - [How to build an AI agent team that ships one deliverable](https://aldena.ai/blog/how-to-build-an-ai-agent-team): 2026-08-10. A working AI agent team needs real roles, a coordinator that only delegates, bounded channels, shared memory, and cost-tiered models. Here is the build.