@FinrepAI
Account based inUnited States
About this account
- Account based in
- United States
- Connected via
- Web
Account-level information from X, not a live location or the device used for a specific post.
AI Copilot for SEC Research
San Francisco, California
Joined May 2024
- Tweets11
- Following52
- Followers32
- Likes13
Finrep AI retweeted
Tushar divulging secret sauce behind the stateless agent runtime at @FinrepAI
As we started building more products at Finrep, we saw that we were building the same agent setup again and again.
So we moved the runtime out of Fina and made it a shared service.
Now, for a new product, we mostly just need to add its tools, skills and prompts. The rest is config.
Here’s how we did it:
Finrep AI retweeted
Staging environment was built for Humans. @tushaaar19 and engineering team @FinrepAI rebuilt it for AI Agents. Here’s how 👇
Our AI agent can now create a test environment, deploy a feature branch, and test it in a browser, all from a Slack request. It sends the preview URL, results, and a video back to the same thread for our team to review.
ALT We spin up a dedicated Kubernetes namespace for every feature request. Our internal coding agent deploys into it, tests the flow end to end, records the run, and posts the video to Slack.
Finrep AI retweeted
Our AI agent can now create a test environment, deploy a feature branch, and test it in a browser, all from a Slack request. It sends the preview URL, results, and a video back to the same thread for our team to review.
ALT We spin up a dedicated Kubernetes namespace for every feature request. Our internal coding agent deploys into it, tests the flow end to end, records the run, and posts the video to Slack.
Finrep AI retweeted
As we started building more products at Finrep, we saw that we were building the same agent setup again and again.
So we moved the runtime out of Fina and made it a shared service.
Now, for a new product, we mostly just need to add its tools, skills and prompts. The rest is config.
Here’s how we did it:
Finrep AI retweeted
Good one from @tryramp team. This is going to be one of the defining problems for enterprises.
Prices from the big labs keep going up, not down. And once you've built on one, switching is expensive, so you're stuck paying whatever they decide to charge. Most teams are already spending millions and can't really say what they get back. For finance, a cost center, that's very hard to justify.
We also started planning a router at @FinrepAI a while ago and have been building it since. It runs on-prem, so your data stays in your environment.
Today we’re launching Ramp Router.
3 years ago, we built an internal LLM router at @tryramp that powers AI products for 70,000 customers.
Back then it was mostly about saving money.
Now it feels obvious: the best model changes constantly. GPT, Claude, Gemini, Grok, Qwen, DeepSeek, Kimi, GLM - prices and capabilities move every week.
So we’re opening up access to everyone.
One OpenAI-compatible endpoint. The right model for every request. Lower cost without rewriting your app.
Reserve access to use it.
Finrep AI retweeted
Dev's observation is correct. The number of US public companies has roughly halved since the late 1990s. Founders now talk about being public as a distraction to be postponed for as long as possible.
I believe being public is a profound act: a company telling its story to the world. This transparency should be the strongest thing about the public markets, and retail investors can participate in wealth creation in generational companies.
Founders now talk about being public as a distraction to be postponed for as long as possible. Let's investigate at what that distraction actually is - communicating with the regulatory body (SEC) & investor community (Wall Street).
With @FinrepAI - I want the reporting burden to disappear as an argument for staying private. Our ambition is to help every company tell its story to the world with the clarity and rigor once reserved for the largest public companies.
I’ve been deeply involved with 5 public companies, including 2 as ceo. Cos typically went public for three reasons. To raise capital, to give early investors & employees liquidity, & to build credibility, because large enterprises preferred doing business with public cos over private ones.
Today, private capital is abundant. Existing shareholders have ways to find liquidity. Enterprises no longer view being public as a meaningful signal of quality.
However, mgmt teams spend enormous amounts of time & energy managing quarterly expectations & stock price volatility. Most of that effort creates little lasting value.
There is a reason there are far fewer public companies today. Unless you need access to massive pools of capital, the case for going public is far less compelling than it once was.
Finrep AI retweeted
Accounting services and related areas in the US is a multi-100 billion dollar industry which has been traditionally dominated by a duopoly model where Native American CPAs combined with offshore services firms in India, Philippines and other low cost destinations.
AI is ready to disrupt this industry and do so in a big way.
The good news is that many Indian-origin founders are building startups in this space which will see billions of dollars outcomes at scale in the next decade.
Few names that come to my mind:
@maxima_dot_ai @maximor_ai @FinrepAI @GrowfinApp
We are still getting started and sky is the limit. There could be many more in the making. 💪🏽
Finrep AI retweeted
Prompt Engineering Is Becoming A Quiet Finance Skill hubs.li/Q044hBV30 Written by @gmisrag of @finrepai
Finrep AI retweeted
~2 years at @FinrepAI
built a lot, but one thing that keeps coming up is this: problems that look solved usually are not.
pdf search was one of those. looks fine until you deal with 1000+ page documents. most solutions are heavy, expensive, or just not built for real product usage.
we stepped back and rethought it end-to-end. got it from ~60 seconds to ~50ms.
wrote about it:
finrep.ai/blog/instant-pdf-s…
Finrep AI retweeted
Why Technical Accountants And SEC Reporting Professionals Should Adapt To AI Early And How To Do It Right hubs.li/Q049CL_m0 Written by @gmisrag of @finrepai
