Coding

AI coding tools, software automation, and the shift from assistant to agent.

Author Johannes Haus
Last updated

This category contains 6 tracked predictions. Each page includes the original claim, current evidence, counterevidence, and what would change our assessment.

200,000–250,000 AI coding agents run in parallel Not Yet Testable
March 2027 Updated 2026-03-13 40% confidence

OpenBrain runs 200,000 Agent-3 copies in parallel. (page 13. Note: Appendix I page 54 states 250,000 copies — an internal source inconsistency.)
Coding agents provide significant real-world value Confirmed
Mid 2025 Updated 2026-05-18 95% confidence

AI coding tools transform software development, becoming indispensable for most professional programmers.
Coding fully automated but research taste still requires humans Not Yet Testable
Early–Mid 2027 Updated 2026-03-13 55% confidence

Now that coding has been fully automated... research taste has proven difficult to train due to longer feedback loops and less data availability.
RE-Bench score reaches 1.3 On Track
Early 2026 Updated 2026-03-13 60% confidence

1.3 on RE-Bench matching top expert humans given 8 hours at well-defined AI research engineering tasks.
Superhuman coder emerges Not Yet Testable
March 2027 Updated 2026-04-27 55% confidence

A superhuman coder (SC): an AI system that can do any coding tasks that the best AGI company engineer does.
SWE-bench-Verified score reaches 85% On Track
Mid 2025 Updated 2026-05-05 90% confidence

We forecast that mid-2025 agents will score 85% on SWEBench-Verified. (page 3, footnote 3. Note: the 72% starting point is tracker-added context, not from the source.)