AI SRE platform to run production reliably
San Francisco, CA
- Tweets 1,329
- Following 270
- Followers 2,602
- Likes 3,092
ResolveAI retweeted
What makes a founding team worth betting on?
For Unusual’s John Vrionis: a technology inflection, deep customer empathy, domain expertise, rapid learning, and the conviction to pursue something fundamentally different.
John has known @resolveai’s @spirosx for nearly 20 years and worked with him and @maynkag across three companies.
We believe their technical depth, humility, and non-consensus conviction make them exceptional founders. We’re grateful to partner with them.
1
1
14
2,036
Today we're launching the Resolve AI plugin 2.0 for Cursor, Claude Code, and OpenAI's Codex. Install it and your coding agent has live production context via Resolve: open alerts, recent deploys, root cause investigations, and remediation plans that turn into pull requests.
4
1
12
929
Read more about the new capabilities with Resolve AI Plugin:
resolve.ai/blog/resolve-ai-p…
1
3
212
LLMs are extremely good at generating large volumes of knowledge, but ensuring that knowledge is useful is where the real value lies.
4
421
Listen to our very own @swaritjoshipura describe why a production agent is DOA if it lives inside a single or limited number of tools, and why Resolve queries your stack live at runtime instead of ingesting it.
2
10
880
Evals. Evals. Evals. Hillclimb. Hillclimb. Hillclimb.
-Dhruv Mahajan, Chief AI Scientist at Resolve AI
Don't fall for the 80% trap. With a production agent, the hard part is the last 20%. You need a loop to keep those models improving, and that loop requires evals built against your own production.
Creating the right evals is a core ML problem: easy to get wrong, easy to make too easy, and hard to know what you should be climbing toward. It's not a project you finish, and most teams aren't staffed or ready for that.
3
1
15
2,134
What does it take to let an agent take action in production?
Listen as Claire articulates the significance of guardrails and governance for agents -- especially those like Resolve that are taking actions in prod.
1
2
8
460
Why are evals on production agents so much more difficult than evals on coding agents?
1
1
27
15,099
We asked you to share a time you "fumbled a 9," and you delivered.
From first-time on-caller horror stories to debugging a SEV2 two drinks in on vacation, one thing was abundantly clear: bagging more 9s ain't easy.
But with Resolve AI, we help make it easier than ever. If you want to learn how, or want to share your own story, check out the link below and we'll send over your very own Bag More 9s bag.
2
2
351
Bag your own Bag More 9s bag:
resolve.ai/bag-more-9s
145
A frontier model can produce a thousand coherent answers. An incident needs exactly one correct one.
That is the hard part of running AI in production. When you generate a line of code or draft an email, any plausible output can work. With an incident though, there's only one root cause. Miss it, and the confident wrong answer is the expensive one.
Frontier models will not fix this. The work is in the harness around the model.
We laid out exactly what that harness actually requires in our full ebook below 👇️
1
3
298
ResolveAI retweeted
very very cool hosting a group of engineering leaders @MLCricket, and hearing the universe boss @henrygayle stop by & talk shop 🏏 w/ @resolveai
1
15
1,345
To build, or not to build, that is the question.
You cannot entrust your on-call rotation to a frontier model chained to a few MCPs. Those who build it themselves meet with answers spoken proudly yet false, invoices swollen past all reckoning, minds that keep no memory from one investigation to the next, and agents let loose upon your production realm, ungoverned and unguarded.
4
5
27
148,010
Come learn where homespun AI for production most often cracks and crumble👇️👇️
resolve.ai/build-vs-buy
2
194
Find us next week in Alamo Sq Park 😉
resolve.ai/bag-more-9s
154
