Who makes the final decision on trades at Arca?

Human advisors do. Agent roles are limited to analysis and proposing solutions, and trading and any other money movement is handled by a human advisor.

Which AI models does Arca use?

Arca uses models from frontier labs like Anthropic and OpenAI, with Claude as its primary model provider. Its own harness automatically selects the best model for a task based on internal benchmarks.

How does Arca protect client data when agents write code?

It translates the agent-written code into a strict expected format, discards the code, and applies guardrails to that format before it reaches the back end.

Large Language ModelsAI Business & IndustryLatent Space3m agoNov 1, 2026, 04:00 JST

Arca's AI agents run up to 10 days, but humans still trade

Send on LINE

3 Key Points

What happened
Arca, which recently raised $48.5 Million led by General Catalyst, built its own agent harness and a per-client memory system; AI Engineer Aman Singh said agent tasks can take up to ten days.
Why it matters
The setup keeps agents to analysis and proposals while advisors handle all trading and money movement, so AI can guide long financial workflows without taking on client risk, Singh said.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Get these free every morning

Context & Analysis

Arca's approach reflects a specific problem in applying AI to wealth management: the work is long-horizon but stop-start. As Aman Singh described, tasks can run up to ten days with many actors, decision points, and handoffs in between. That is why the firm built an internal memory system that tracks, for each client, what has happened, what is being tracked next, what needs to happen in the next couple of days, and which individual is responsible for which action.

The firm's choice to own its harness rather than use a managed one is tied to flexibility. Because the harness automatically selects the best model for a task using Arca's internal benchmarks, owning it means Arca can change model providers if needed. Arca currently uses models from Anthropic and OpenAI, with Claude as its primary provider.

Safety is handled through deterministic restrictions that take decision-making away from the agent. Agents can execute some tools and tasks themselves, but not ones that involve risk to the client; trading and money movement stay with human advisors. Arca also runs verification checks on known AI security issues such as prompt injections, and uses sandboxes for computation with extra measures for sensitive data. Separately, a meta-agent analyzes advisor feedback and agent traces to suggest harness improvements.

FAQ

Who makes the final decision on trades at Arca?: Human advisors do. Agent roles are limited to analysis and proposing solutions, and trading and any other money movement is handled by a human advisor.
Which AI models does Arca use?: Arca uses models from frontier labs like Anthropic and OpenAI, with Claude as its primary model provider. Its own harness automatically selects the best model for a task based on internal benchmarks.
How does Arca protect client data when agents write code?: It translates the agent-written code into a strict expected format, discards the code, and applies guardrails to that format before it reaches the back end.

Latent SpaceRead Original Article

Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Get it free

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.