FutureSearch Blog
futuresearch.ai · Company blog · on HN since 2024
FutureSearch, an AI forecasting startup, writes about LLM economics, OpenAI analysis, and security incidents affecting Python packages.
What HN saysHN links it for its report on the LiteLLM PyPI supply chain compromise and its analysis of OpenAI Deep Research failures. Its team also posts job listings.
18HN people shared it
5linked it in comments
39stories submitted
238story points
2 months agolast shared
2024people per year2026
Most-shared posts
- 1LiteLLM PyPI has been compromised an hour ago, do not update3 people · 17 points · first shared 2026 · 3 comments on HN
- 2Estimated OpenAI Revenue Breakdown3 people · 5 points · first shared 2024 · 1 comments on HN
- 3Show HN: FutureSearch, AI forecasting you can verify2 people · 13 points · first shared 2024 · 1 comments on HN
- 4OpenAI Deep Research – Six Strange Failures2 people · 8 points · first shared 2025 · discussion on HN
- 5Unit Economics of LLM APIs2 people · 3 points · first shared 2025 · discussion on HN
- 6The Karpathy Interview, 6 Months After AI 20271 people · 38 points · first shared 2025 · 24 comments on HN
- 7How long until AI automates all cognitive labor?1 people · 31 points · first shared 2026 · 34 comments on HN
- 8Apple's plan to power Siri with ChatGPT was a predictable failure1 people · 15 points · first shared 2025 · 1 comments on HN
- 9How AI Finds Fuzzy Duplicates in Large Datasets1 people · 10 points · first shared 2026 · 1 comments on HN
- 10LLM Agents Solve the Table Merging Problem1 people · 10 points · first shared 2026 · 1 comments on HN
- 11Estimating the market value of top AI researchers1 people · 9 points · first shared 2026 · discussion on HN
- 12How Predictable Are the Oscars?1 people · 8 points · first shared 2026 · 2 comments on HN
- 13Higher effort reduces deep research accuracy for Gemini Flash 3 and GPT-51 people · 7 points · first shared 2026 · 1 comments on HN
- 14Agents Sometimes Catastrophize1 people · 7 points · first shared 2026 · 1 comments on HN
- 15Some rare examples of AIs being underconfident1 people · 6 points · first shared 2026 · discussion on HN
What HN people wrote when they linked it
ddp26 · 2025-01-03
FutureSearch | Software Engineer, Research Engineer, Research Scientist | Full Remote (global) | Full-time
FutureSearch (https://futuresearch.ai) is a seed-stage startup building AI systems that reason about the future reliably.
We're an elite 7 person team, who all worked together prior to FutureSearch. My co-founder is the former Head of AI at Metaculus, and I'm the former CTO of Metaculus & creator of Google's currently running internal prediction market.
We're backed by top AI investors, and we're building systems evaluated by the hardest, truest eval possible: predicting the future. We have sold our outputs to Google, OpenAI, and Amazon.
If you're fascinated by what LLMs can and can't do, have excellent technical and reasoning skills, and want a competitively-paid, fully-remote job, please apply!
See the jobs and apply at https://futuresearch.ai/careers.
Bullhorn9268 · 2026-03-25
I am from futuresearch and went through this with Callum (the OG). We did a small analysis here: https://futuresearch.ai/blog/litellm-hack-were-you-one-of-the-47000/ of the packages and also build this mini tool to analyze the likelihood of you getting pwned through this: https://futuresearch.ai/tools/litellm-checker/
kevml · 2026-03-24
More details here: https://futuresearch.ai/blog/litellm-pypi-supply-chain-attack/
cratermoon · 2025-02-20
meanwhile: OpenAI Deep Research - Six Strange Failures: <https://futuresearch.ai/oaidr-feb-2025>
incorrecthorse · 2025-02-18
> So, what exactly has generative AI actually done? Where are the products?
The product is ChatGPT, actually.
If LLMs are a bubble, then you should expect most of OpenAI's revenue to come from its API (which is used by startups which have raised money to do "magic AI stuff", and the bubble would pop when investors would stop giving the money). But according to https://futuresearch.ai/openai-revenue-report, revenue from the API accounts only for 15%, the other 85% being the different subscriptions offers, including 55% of ChatGTP Plus subscriptions -- that is, _direct consumers_.
This doesn't prove that it isn't a bubble (the consumers could realize it's useless and then leave some time later), but it makes it less likely IMO.