Salesforce ships Koa, its first reasoning model, born from a 3-year-old function calling bet
huan__wang · x · 2026-09-18
A researcher on the Salesforce AI Research team recounts how, almost 3 years ago, a small team bet that teaching models to take the right actions — function calling — would matter enormously for AI agents. That bet shipped this week as Salesforce Koa, the company's first reasoning model. The thread traces the journey from the early call to launch, arguing tool-use capability is foundational for agents.
More from Companies & People
- Essay: Three years into AI, the fork between human-AI symbiosis and parasitism — yangyi · 2026-09-18
- 90-minute interview captures Sequoia legend Michael Moritz; Jim Goetz recalls being mentored by him — GuptaRK22 · 2026-09-18
- First-ever PyTorch Day Japan lands in Tokyo on December 10, CFP open till Sept 27 — PyTorch · 2026-09-18
- NumPy creator Travis Oliphant on funding open source foundations in the AI era — teoliphant · 2026-09-18
- Glean beats Claude in customer eval: 30% cheaper with better quality, saving $500K a year in tokens — verrsane · 2026-09-18
- Astra for Law posts big legal benchmark gains as Mollick asks if labs will eat every AI vertical — emollick · 2026-09-18