™Using Agents for Long-Horizon Financial Work — Aman Singh, Arca (26:05)
Latent Space interviews Arca head of AI Aman Singh at AI Engineer New York, 26:05, on Arca's custom agent harness, multi-day task memory, and safety limits in financial services.
Key points Tap a time to jump
How AI supports human financial advisors, with advisors making final decisions
Choosing models with internal benchmarks
Building a custom harness and routing tasks
Handling a $100,000 tax bill: agents and human approvals
Agent memory across long-running tasks
Safety guardrails and prompt injection defenses
Singh's line: agents prompt people more than people prompt agents
Why trade execution stays with human experts
Client examples: hidden fees and proactive service
Comments
More to watch
All →- 27:34Building ChatGPT for Finance — Joseph Kim, OpenAI's Head of Financial Services (27:34)
- 55:38WHOOP CEO Will Ahmed: 50B Hours of Health Data, a 4 Million Member Data Moat, & the FDA Fight (55:38)
- 1:16:38Why Agents Love Supabase — with Paul Copplestone and Ant Wilson
- 2:27Genie 4: A new generation of world simulation
Latest news
All →- Nov 1Family Sues OpenAI, Alleging ChatGPT Guided Epilepsy Patient Off Medication Before Her Death
- Nov 1Meta Built Humanoid Prototype Max With Fourier in Three Months, Reports Say
- Nov 1OpenAI to Pool ChatGPT and Codex Usage Into One Allowance From December
- Nov 1OpenAI Tightens ChatGPT Violence Safeguards: Teen Weapons Limits, Attack Detail and a Six-Month Safety Report, Explained
- Nov 1S&P Cuts Meta's Outlook to Negative on AI Spending and Cash-Flow Uncertainty
- Nov 1OpenAI Tightens ChatGPT Violence Safeguards: Teen Weapons Info Restricted, Safety Report Every Six Months From 2027
Today
Nov 1 Sunday- Brief
- GitHubGitHub Weekly Vol.3: 6 Projects Worth a Look This Week
Tested or read — weeklyAll →
- Quiz
- Call