Same Research, Sharper Wedge: Why Rippletide Now Leads with Write-Access Safety

If you visited rippletide.com this week, you noticed the change. The site used to open with our category: the Decision Runtime for enterprise AI agents. It now opens with a promise: let AI agents write to production, safely.

This is not a new company. It is the same research, packaged where the pain is sharpest. Here is the honest story.

What we kept hearing

Over the last year we deployed our decision technology with enterprise teams and agent builders, won the OpenAI Codex hackathon in San Francisco, and joined Station F's F/AI program. Across all of it, one pattern kept repeating.

AI agents read very well. They draft, recommend, retrieve, summarize. The blocker never was intelligence. The blocker appears the day the agent needs to write: issue a refund, close a P1 incident, update a CRM field, cancel an account. That is the moment an enterprise security review steps in and asks questions most teams cannot answer with evidence: was the right context present? Did the action match policy? Who approved it? Where is the trace?

Deals stall right there. Not on model quality. On write access.

What changed

We packaged our technology to attack exactly that blocker:

• A Free Risk Review: one 30-minute session to map an agent's tool surface and rank its riskiest write-actions.
• Safety Packs for support, ITSM, and revenue agents: risky write-actions, evidence checks, approval rules, unsafe scenario tests, and audit-ready traces, built and proven on your agent in 10 working days.
• A security review dossier: the structured proof an enterprise security team needs to say yes.

The website now leads with this, because it is the fastest way to be useful to a team blocked at the write-access door today.

What did not change

Everything underneath. Rippletide started as deep research: a hypergraph decision database that combines memory and reasoning, automatic ontologies that turn policies and workflow logs into explicit rules, neuro-symbolic evaluation that produces deterministic, traceable decisions. That research did not go anywhere. It is what runs a risk review, builds a Safety Case, and previews an approval decision. It is also why our long-term direction is unchanged: as Safety Cases are proven, they expand into runtime control of high-risk agent writes in production.

If you followed us for the research, it all still lives here: Automatic Ontologies, the Context Graph for agents, and The Decision Layer podcast.

Why we are telling you this

Positioning changes can look like amnesia from the outside. This one is the opposite: it is two years of work finding its sharpest expression. The research earned the right to be packaged; the packaging makes the research useful in ten working days instead of a nine-month platform conversation.

If your agent is stuck at the write-access door, book a free risk review. And if you want to argue with us about decision databases, the door is open too. Same team, same conviction, sharper wedge.