Iacon operates Optimus, the first autonomous research scientist.
Bring an agent or model and a goal. Optimus researches algorithms, designs experiments, and
evaluates what improves it.
Website · Research · Discord · X · LinkedIn
We believe superintelligence will not be a single monolithic model, but a network of domain-specific models: language models, world models and RL policies. If an agent can practice, observe and improve within an environment, it can be trained. We are automating this universal loop.
| Optimus | An autonomous RL engineer. Set the environment, rewards and success criteria; Optimus runs data, RL environments, reinforcement learning, fine-tuning, post-training, benchmarking and evaluation end to end, and decides what to try next. |
| Arkenos | The research workspace Optimus runs in, natively on your own hardware. |
| Nenyax · open source | One contract for every RL environment. Load any format (Gymnasium, OpenEnv, Verifiers, NeMo Gym, Harbor, reasoning-gym), judge it your way, and improve any model with any learner. pip install --pre nenyax |
Nenyax is Apache-2.0 and built to be extended: drivers, sandboxes, model providers, hubs, learners, judges and telemetry are all plugins. Companies and researchers are welcome to bring their own; see the contributing guide.
Research grants: a free year of Arkenos for independent researchers and early-stage teams, details on iaconautonomics.com. Contact: dev@iaconautonomics.com
