Ben Roesch

Co-Founder/CTO @CultivateLabs. Builder of @arcanalysis, the Analyst Research Companion and @raif_ai, a Ruby/Rails engine for LLM apps

[Cultivate Labs @cultivatelabs]

Claude Fable 5 launched June 9, suspended worldwide June 12 under a U.S. export-control order. Will it be restored for U.S. users before Aug 1, 2026?

Our AI + human forecast: 57% Yes. Track it on Hinsley:
https://t.co/qcSt78W9cA

[Cultivate Labs @cultivatelabs]

New in Hinsley: multiple choice forecasting questions (beyond Yes/No) and crowdsourcing campaigns - a survey-like experience to collect predictions from external participants via shareable links.

https://t.co/qZYrkiKpS0

[Cultivate Labs @cultivatelabs]

AI forecasting isn't just about accuracy - it's about agility. Now, the questions you're asking can evolve as fast as the world they're trying to anticipate. https://t.co/45ZvkcYNwD

[Cultivate Labs @cultivatelabs]

For several months, we've had Hinsley (https://t.co/px7hAOwVuD) submit regular forecasts on one of the forecasting platforms we run in conjunction with Bertlesmann Foundation (https://t.co/JnNTMjokjM). As of today, HinsleyBot is officially at the top of the accuracy leaderboard!

[Ben Roesch @bcroesch]

v1.4.0 of @raif_ai is available now for your #rails apps. Google adapter, plus a long list of agent, web admin, & general quality-of-life improvements. Full list: https://t.co/1QIqXVnaa2

[Ben Roesch @bcroesch]

My latest blog post from the @cultivatelabs blog, spelling out why the obsession with forecasting accuracy risks ignoring the bigger challenges - https://t.co/sIv29CrjB6

[Ben Roesch @bcroesch]

@aymorgan @c_sonnier Raif v1.3.0 just shipped with built-in eval support. Might be of interest! https://t.co/feWkR49nKa

[Ben Roesch @bcroesch]

@raif_ai @mattsolt @cooperx86 @lucianghinda might be of interest!

[Ben Roesch @bcroesch]

Just shipped v1.3.0 of @raif_ai with built-in support for LLM evals! Evaluate the quality of LLM responses for different models in your Ruby on Rails apps. Built-in support for using LLM-as-judge. https://t.co/feWkR49nKa

[Ben Roesch @bcroesch]

@windlejacob12 We don't use langchain (we extracted @raif_ai from our app), but we've do most of our orchestration via a Workflow model/class. Largely just a pre-defined series of LLM calls. We tried some more agentic stuff but found workflows to work better.

[Raif AI @raif_ai]

Raif has a new docs site! https://t.co/AdnKhGwl8k

[Ben Roesch @bcroesch]

We released a new version of Raif, a Rails engine for working with LLMs. It now supports streaming responses in chat & working with provider managed tools like web search/code execution. https://t.co/lqInNIdK0J

It powers all of the LLM features in ARC: https://t.co/qyR2YfYJoR

[Ben Roesch @bcroesch]

@_samadahmed @neejkapoor Raif is a Rails engine that includes a Chat UI -- https://t.co/lqInNIdcbb

[Cultivate Labs @cultivatelabs]

🎊Our 2024 wrap-up just hit inboxes!

Dive into our CEO @amsiegel's year-end review + get early access to ARC, our new tool combining human expertise with #AI capabilities to help analysts work smarter.

Take a look: https://t.co/2c3oBK63bK

[Cultivate Labs @cultivatelabs]

Our March #newsletter shares several recent highlights, including:

🤝a partnership with @RANDCorporation & @INFERpub
📊a workshop during a @BertelsmannFdn transatlantic relations event
🏛️2 experts' take on forecasting for national security

Read more >> https://t.co/6fKiFE58nK

[Cultivate Labs @cultivatelabs]

How does #crowdforecasting work, you ask?

Our latest blog dives into:
✔️a brief history of forecasting
✔️our typical outputs for analysts and decision makers
✔️how you can harness crowd forecasting for your organization

See more >> https://t.co/OiMo8aWZ6L

[RAND Forecasting Initiative @RANDforecasting]

We are pleased to share that #INFER is now led by @RANDCorporation. We look forward to this new chapter that will enable INFER to leverage #RAND expertise in our mission to advance the #forecasting capability of the US govt.

See our announcement>>https://t.co/jV5BcxT9EB

[Cultivate Labs @cultivatelabs]

Having trust issues with the news these days?

Our latest blog shares how crowdsourced #forecasting is cutting through #disinformation to reveal truth.

Read more >> https://t.co/roImmPuUM7

[Cultivate Labs @cultivatelabs]

An effective #crowdforecasting program hinges on asking the right questions. How does Cultivate develop forecast questions that are useful and actionable for decision-makers?

Learn more about our process from a Q&A w/ our teammates in our latest blog >> https://t.co/MHiz5mvntQ

[Cultivate Labs @cultivatelabs]

Does forecasting accuracy really matter?

Yes, but improving accuracy isn’t necessarily the most valuable aspect of a crowdsourced forecasting effort – and it shouldn’t be its main focus.

Learn more in our latest blog >> https://t.co/SbxQTq9TuT