Agentic Development Future and Current Directions - Nov 2026

Here are things I'm seeing from across the web - that I think will be big in 3-12 months. Just putting it here so I don't forget. Could be useful for you as well.

1. Formal Software Verification (TLA+ and others, for agent-based work)
2. Agent Experience Design and Testing
   • (l.a AJX)
3. Single-user and multi-user long-lived agents (like OpenClaw, Kiro Crew)
   • Security
   • Privacy
   • Context isolation
4. Sandboxing for agents (not new but gaining)
   • Types
   • Best and worst practices
   • State of the art
5. Factory Engineering (how to design good software factories; the future of CI/CD in the age of agents) (ie my factory post but for software)
   • This covers many smaller fields:
     1. Types of agents and their roles (chief of staff, agent per task, temporary vs long-lived agents)
     2. Role of testing in factories: which types of tests make sense
     3. Environment management in agentic pipelines
     4. Company Brain, context engineering, security, privacy
     5. Sandboxing
     6. Long-lived vs ephemeral agents
6. Company Brain (not new but gaining)
   • Types of brains and contexts
   • RAG vs graphs
   • Tools like RepoSwarm
7. Agent Swarms: cross-agent communication and interaction
   • Like Grok bots, which can create and talk to other bots easily
   • Tools like Kiro Crew bots and Grok bots
8. ProdoTyping
   • SDLC for prototyping (Lovable, Base44: many orgs are moving to self-prototyping flows)
   • Tools like LowKey plus an auto-vending control plane
9. Multi-agent code conflicts
   • How do 1000+ agents, with 1000+ commits on the same repo, avoid bottlenecks?
     • Still unsolved
10. Decision Models
    • Best and worst practices
    • How to build or train your own decision model
11. Local Models (not new but gaining)
    • Used together with remote models
12. Harness Engineering (been going on for a while)
    • Learning and using harnesses like Pi
    • How to build your own harness, or extend Pi and others