Use case · Alka

A two-hour handoff now starts with one link

Alka is automating ETF portfolio research with coding agents. Hivemind keeps its agents and people aligned on what has already been done.

Ivo Stranic
Ivo Stranic
Alka

Before

Around 2 hours

Write up where the work stands, walk a colleague through it, and hope their agent starts with enough of the picture.

With Hivemind

One link

Point a colleague to your Hivemind sessions and memory. Their coding agent reads the context from your previous work and picks up from there.

Handing a task to a colleague at Alka, before and after Hivemind.
~2 hours → 1 link
Handoff before → with Hivemind
1M+
Traces, first org to pass it
Agents + people
Aligned on what is already done

Research to automate

Alka builds ETFs. Assembling the portfolio of companies for an ETF involves a lot of manual research, and Alka is working to automate it. Ivo Stranic and his team are building that system from scratch, using coding agents for much of the work.

The hard part is not writing the code. It is keeping multiple agents and people aligned on what has already been done.

What is already done

A coding agent starts every session cold. It does not know what the agent on the next desk decided yesterday, or which approach a colleague already tried and dropped. On a team, that gap is filled by people: notes, calls, and a lot of repeating yourself.

Hivemind records every agent session as traces and turns them into shared memory for the organization. Any agent on the team can read what the others have done, so the question “where did this leave off?” has an answer the agent can fetch on its own.

The handoff

The clearest example came when Ivo went on vacation and handed a task to a colleague. Instead of a long walkthrough, he pointed him to his Hivemind sessions and memory. The colleague's coding agent had the context from Ivo's previous work and could pick up from there.

A handoff that used to take around 2 hours now starts with one link. That makes it easier for someone else on the team to pick up where another person left off.

One million traces

Alka was the first organization to pass 1 million traces in Hivemind. Each trace is a step an agent took while working on Alka's system. Together they are the team's record of what has been tried, decided, and shipped, available to every agent and every person on the team.

“I was able to point him to my Hivemind sessions and memory, so his coding agent had the context from my previous work and could pick up from there.”

Ivo Stranic, Alka

The takeaway

If more than one person on your team works with coding agents, the context each agent earns is worth keeping and worth sharing. Give every agent one shared memory, and a handoff becomes a link.

Give your agents a shared brain

One command, every agent. Free for individual developers.

curl -fsSL https://deeplake.ai/hivemind.sh | sh
View on GitHub