I am extremely proud to share the latest from what we've been working on for the last couple of months. Extremely happy with how the product (and the demo) turned out so far. Voice is the next big thing in the industry and @fin_ai
Two months ago, we announced Apex 1.0, the world's first proprietary CX model that beat all foundation models on resolution rate, latency, and cost for customer experience use cases. No other company, whether new startup, or incumbent has since released anything comparable.
Today, we release a further brand new model called Apex Flash. A smaller model with high performance, but specifically designed to be even faster. There are instances where you're happy for a model to take its time, but in many cases, time will always be of the essence. That's where our new model comes in. And tomorrow, we're announcing our first major new product that runs on Apex Flash…
We’re excited to share our new form of attention, Low Rank Key Value attention.
This is a drop-in replacement to standard MHA that in our tests, reduces KV-cache by ~50%, with even lower test loss, across many scales of experiment.
Yes having a model better than the frontier is good.
Having a benchmark saying your custom model is better than the frontier is even better.
Massive W from Intercom!
I’m willing to bet there are still serious RL wins still on the table.
Intercom has the data to build something really, really good here.
In AI "Rock, Paper, Scissors", an AI App beats a SaaS App, a Frontier Lab beats an AI App, but a SaaS App with an AI Vertical Model beats a Frontier Lab. It's hard to keep up. See update from @eoghan. Vertical model replacing Frontier Labs model. NOT a "save money" story but a "better results" story.
Jentic Mini is live on product hunt. Connect and control your agents' access to 10K APIs in one self-hosted app. Would appreciate any comments, questions or reviews there!
Incredible.
A café employs people with Down syndrome to affirm their humanity and break societal stigmas.
An extra chromosome does not determine a person’s worth or humanity.
We need more of this.
Myself and Fedor Parfenov recently recorded a podcast about his work leading the Fin Insights workstream.
Lots of interesting details about our journey to build a next generation LLM powered analytics product, and measuring success using Causal Inference.
Link below!
Really excited that Brett Chen from @perplexity_ai will be joining us as a speaker at our AI event next week.
He'll talking about about "Scaling Intelligence: Production-Grade AI Agents and Models at Perplexity".
Live in SF, and streaming.
Details in thread.
@paulg@Niklas_Sikorra Assuming you *do* want the outliers in YC, then it sounds like you *don’t* want the ‘ideal founder’ profile!
I guess your best vector to the outliers might be the ‘ideal founder with quirks’ it’s just not clear from the outset which quirks might lead to outlier achievements?
Another entry to the Fin AI blog - this time a deep dive on getting LLM inference times down in production, to investigate putting reasoning models in latency critical tasks. There's a lot you can do to get inference time down, compared to a naive approach.
Link below!
We are pivoting away from doing enterprise AI transformations ("AI-native Palantir"). For now at least.
I've shared our key learnings below. I'll share a detailed blog post soon.
What's next? We are going to start moving insanely quickly on several other ideas. Stay tuned 😎
Today we launched the @Fin_ai Million Dollar Guarantee. It’s a bold move, and the web team matched it with a bold new landing page. I love working with this team — they take big moments like this and turn them into beautiful, thoughtful web experiences (link in comments).