We just hit an important milestone with OpenNeko autonomous AI agents. 90% of the Cybersecurity Maturity Model Certification (CMMC) controls assigned directly to the product are now fully shipped. ✅
https://t.co/aKCv1Mb5wr
a new benchmark to evaluate agents in complex organizations with large multiple databases, code, providing business insights, reacting to changes, etc. still working on a better name for the benchmark "DeepOrg" is what i'm thinking.
harnesses matter with graphjin + it's embedded ax agent (RLM) even the cheap fast gemini-3.5-flash-lite scores 84%
https://t.co/8Mnz81U5Hc
In the begining, people at your client will want you gone, except your project sponsor. So its not an honest feedback loop, they want you out of there ASAP.
Your job is to manifest the CEO or CXOs vision when everyone else is telling them it cant be done.
So not just tech skills, you need to be able to sell a vision to a team of people who dont trust you, and you need to be able to win their confidence and carry them along on the ride.
Introducing Apps in the OpenNeko Agent Harness
Just launched apps in the openneko agent. It lets you create your on prem versions of
Salesforce , Zendesk, other expensive SaaS, with one prompt. Manages changes as well. Then import your data and let the agent loose on your newly liberated data.
Ask complex questions about how you are doing as a company, and let the agent run watchers for things that happened and that did not happen, and even take corrective action.
@mstrazds@PrimeIntellect Also check out axllm and graphjin. And the RLM agent that you can turn on within graphjin. Graphjin is a game changer for data access. Have deployed over a 5000 table snowflake cluster with billions of records. Answers match 100% with ground truth in custom evals
@rohanpaul_ai Yup. There are implementations of https://t.co/W7VMjfcuoX. the operational layer of ai that is personalised to a company. And its open source.
@zotspot@Eminemead It automatically figures out the reorder point from your database. Unless of course that data doesn't exist then you'll need to tell it explicitly
@zotspot@Eminemead It's already being used in production. Just chat normally and tell the ai "watch for stock levels of x dropping below reorder point". And it will setup a continuous watcher to do that. Doesn't consume any LLM api until the event happens and it needs to act on the event.
@Eminemead I've been contemplating gbrain. There is a durable memory in there that is basically a similar implementation (file based) but also backed by git and pg_vector (similarity search) but I'm going to take another look at gbrain
@yashwanthsai29 2026.. hmm you should have built agent frameworks in 2023-2024 and the harness in 2025-2026. Anyway, not too late. Feel free to give openneko a spin. It harnesses the Hermes agent
@Eminemead I'm hoping to do a series of posts on how some people have been using some of the features. A classic use case is watch stock counts of critical raw materials in a factory so if they drop below reorder point the agent can alert the operator.