Gary Marcus challenges dev: show symbols in your agent harness or you're confused
GaryMarcus · x · 2026-11-16
Gary Marcus pushed back on developer @OwariDa's claim that modern agent harnesses are effectively neurosymbolic, saying he'd like to see a harness that actually adds symbols — otherwise the dev is "confused about the history of the arguments." @OwariDa had argued that "code mode" implementations are just a subset of tool calling (parse LLM output, act, feed results back), not a requirement for SoTA agent performance, citing his own early harness built for GPT-3 and the first ChatGPT release. A notable spat over what neurosymbolic AI actually means in today's agent stacks.
More from AGI Musings
- Debate: Was Tool Calling an Unpredictable Leap, or Predictable Since 2022? — OwariDa · 2026-11-16
- Dean Ball: top AI models mode-collapse into 2020-era governance discourse — deanwball · 2026-11-16
- JOBhakdi pushes back on 'Musk just got lucky with money' view of innovation — JOBhakdi · 2026-11-16
- The AI doom market sells psychological certainty, not probability — YogeshMalik · 2026-11-16
- Runway CEO: AI's history is an ideology of 'everyone gets to make' — c_valenzuelab · 2026-11-16
- The Autonomy Gradient: a framework for how much to delegate to AI — DrKavner · 2026-11-16