Research
Research by Vikas Goel
Independent AI research, published under the ThinkerWave programme. Focused on self-evolving agent systems, evaluation frameworks, and the engineering reality of LLM-based AI at production scale.
Publications
Research paper · ThinkerWave Research · April 2026
Die to Evolve: When AI Agents Mutate Both What They Are and What They Seek
Introduces dual-space optimization for self-evolving AI agents — searching the space of solutions and the space of evaluation criteria simultaneously, with complete identity replacement between generations and epistemic persistence carrying knowledge forward. Validated across five domains. Patent application 202611044024 (Indian Patent Office).
- Self-Evolving Agents
- Evaluation Gap
- Dual-Space Optimization
- LLMs
- Patent Pending
Ongoing & upcoming work
The ThinkerWave research line continues. Areas under active investigation include cross-domain criteria transfer between agent lineages, the boundary conditions of identity replacement in safety-critical deployments, and the relationship between evaluation evolution and verbal calibration in LLM-as-judge architectures.
For research collaboration enquiries, get in touch.
Find Vikas Goel's full body of work at Work & Research, or read more on the blog.