? Telescope AI

?Life - and markets - are inherently unstructured and generative AI unlocks a more adaptive approach to investing by allowing investors to intuitively navigate complexity rather than relying on rigid "store-and-retrieve" pre-defined strategies.

We've integrated Claude 3.7 Sonnet into Telescope. The model shows notable improvements in reasoning, which our initial benchmarks across compliance capabilities. As always, we remain model-agnostic and focused on bringing our customers the most effective AI capabilities, planning for both redundancy and highest-quality output.

Palantir CTO's push for adoption and experimentation: those building and experimenting with AI today are learning exponentially faster than those still analyzing from the sidelines.

For those interested in the best insights into AI, highly recommend watching this fascinating discussion: "DeepSeek, China, OpenAI, NVIDIA, xAI, TSMC, Stargate, and AI Megaclusters" on Lex Fridman's YouTube channel.

This rings true.?Small teams, big impact. Sometimes the most powerful solutions come from the tightest-knit groups.

Just watched Zuck on JRE discussing what's next for AI. Instead of single responses, next-gen AI maps thousands of possible paths before choosing its route. It's not about getting one answer anymore - it's seeing all the branching possibilities and what they mean. Pretty crucial in markets, where one move ripples through the whole system. This is precisely why we keep Telescope model-agnostic - AI moves quick, we need to move with it. Zuck makes a good point about guardrails though. The smarter these systems get, the more careful we need to be with how we build them.

New product launch just in time for the holidays - a touch of green & red! ?? Introducing Chart Patterns: AI-driven trading pattern recognition. What made building Chart Patterns fascinating was the LLM's co-engineering with us: we built a benchmarking platform that ran over 1 million synthetic trading simulations and had AI actively feed back into our detection algorithms. Rather than passive testing with humans making judgement calls, deciding what a "double top" was, we let LLMs both evaluate the results and suggest real-time algorithmic improvements. This build was also grounded with knowledge sources such as Algorithmic trading ebooks and of course foundational knowledge. None of this was possible 6 months ago! https://lnkd.in/geuKW5SE

Our goal: to make everyone's financial journey a little less isolating and a lot more insightful

Todays 4 hour long OpenAI outage - reaching even enterprise-level APIs was a sobering reminder of how deeply reliant we’ve all become on LLMs. At ? Telescope AI we have planned for these events and had redundancy in place but our mitigation methods didn’t go to plan… After a critical hotfix was deployed, all of our core products shifted over to Anthropic.. but globally we were one of many doing the same. Within the hour our platform experienced slowdowns across our core product Ripple—nearly 50% slower. A lot was learnt today!! For us we’re thinking deeper on redundancy, we’ll be evaluating other emerging solutions like Grok and Google’s Gemini - to diversify our stack and minimize future disruptions. We’ll be co-locating embedding models and leveraging our existing redundancy methods for seamless fallback. The incident highlights the reality that generative AI has quickly become critical infrastructure for many organizations. Imagine when we have LLM reliant robotics, personal agents, core infrastructure leaning on LLMs for inference? As model parameters continue to grow, local hosting becomes less feasible. The new reality will demand more robust strategies for resilience, failover, and architectural innovation. Oh.. but when AGI arrives, it’s time for a good book and a cup of tea ??

Epic day celebrating a huge year with the crew @ ? Telescope AI! Crushing it from nature hikes to ocean swims to dining and boat partying - because this team never does anything halfway!