2025: AI agents

This is the year the job title started to change. Instead of building models, the repositories start building workers: software that reads, decides, uses tools and finishes a task. 18 repositories, 10 of them private, 9 of those agency work.¹

What happened

• OpenAI Agents SDK. A customer-support agent in February, a meeting-booking agent in November, math and calculator agents in November.²
• First agency work at Canz Marketing. From May, the agency's first agents repository; from August, a client platform (backend and frontend) and the plant-identification app rebuilt as three services, one of them a chat service on the OpenAI Agents SDK. All private.
• Agent infrastructure. StoreLite IMS in December: an inventory and billing system built in phases, with FastAPI, MCP and agents working together in one backend.²
• Learning alongside. A fork of the Panaversity agentic-AI curriculum in July, and a robotics-textbook hackathon repository in December.

The public repositories

Private work

Ten private repositories: the agency's first agents repository, a client platform in two parts (backend and frontend), the rebuilt plant-identification app as three services (identification, chat, search), a portfolio site for a client, a testing sandbox and a GPT-based agent experiment.¹ They are counted here and never named.

What this era taught

The leap from AI that answers to AI that acts: tools, guardrails and bounded loops.²

Next: 2026: AI employees and MCP connectors.

Footnotes

1. GitHub API, repositories owned by rrizwan98, read 26 August 2026: 18 repositories created in 2025 (8 public, 10 private). ↩ ↩²

2. From Predictive AI to Forward Deployed Engineer: My 6-Year Journey, Raza Rizwan, July 2026. ↩ ↩² ↩³