For those interested, ephemeral graphs in production are thing. @Bauplan_labs were one of the early adopters of @kuzudb and are running 50K+ ephemeral graphs in Kuzu daily for their query planner! 🤯
For those interested, ephemeral graphs in production are thing. @Bauplan_labs were one of the early adopters of @kuzudb and are running 50K+ ephemeral graphs in Kuzu daily for their query planner! 🤯
As someone who works in open source, and engages with other open source devs (maintainers, contributors, users) every day, I think a lot about network effects.
In the end, long term usage and adoption of an open source tool totally depends on network effects. The quality of the…
Recently, @YourTechBud and I have been deep diving into the code base of DSPy, and we're finding different ways to demonstrate the power of signatures, modules and optimizers. The "DSPy way" is a little alien at first (because you give up "control" over what prompts you write).…
@tech_optimist@ThePeshwa@DSPyOSS GEPA works in iterations, where it uses the training set to perform the reflection, and the validation set to choose the next candidate to improve (the pareto frontier tracking) as well as to choose the overall best performing candidate at the end.
Given this, my experience has…
I really like the explanation I just got from @rjurney about how overfitting can occur. "At the limit, it could just write instructions to categorize every instance separately. That would be overfitting". Looking at the verbosity of prompt #3 in this example, it does seem like…
I really like the explanation I just got from @rjurney about how overfitting can occur. "At the limit, it could just write instructions to categorize every instance separately. That would be overfitting". Looking at the verbosity of prompt #3 in this example, it does seem like…
The right question is not to ask if DSPy is production ready, but rather:
>> why would you want to go to prod without a declarative, eval-driven, composable framework? <<
the nightmare isnt a fast evolving framework ... its a repo full of brittle prompt engineered programs
The right question is not to ask if DSPy is production ready, but rather:
>> why would you want to go to prod without a declarative, eval-driven, composable framework? <<
the nightmare isnt a fast evolving framework ... its a repo full of brittle prompt engineered programs
Try out the new graph algorithm tutorial on the Nobel laureate mentorship graph in @kuzudb! In writing this tutorial, I learned the name of a new laureate, Adam Wiess, who won the 2011 Nobel prize in Physics for his work in cosmology. Adam Wiess is the Physics laureate in this…
Try out the new graph algorithm tutorial on the Nobel laureate mentorship graph in @kuzudb! In writing this tutorial, I learned the name of a new laureate, Adam Wiess, who won the 2011 Nobel prize in Physics for his work in cosmology. Adam Wiess is the Physics laureate in this…
Machine & LM cost << human labour cost. In the vast ocean of prompt space, why be a human floating around like a mariner without a compass? Be like @DSPyOSS 🛥️⚓️
Machine & LM cost << human labour cost. In the vast ocean of prompt space, why be a human floating around like a mariner without a compass? Be like @DSPyOSS 🛥️⚓️
This post dramatically undersells what they did here.
Used a couple prompt optimization algorithms to end up with
1️⃣ a prompt that works on a smaller/cheaper/faster model,
2️⃣ successfully processes all problematic inputs,
3️⃣ and the prompt is significantly _shorter_
This post dramatically undersells what they did here.
Used a couple prompt optimization algorithms to end up with
1️⃣ a prompt that works on a smaller/cheaper/faster model,
2️⃣ successfully processes all problematic inputs,
3️⃣ and the prompt is significantly _shorter_
11K Followers 2K FollowingPrefer common sense over hype. Employed @marimo_io, builds https://t.co/KOwpfLMpuQ, https://t.co/EZ7HMX8u9J, blog https://t.co/zREZ8H2IBZ.
387 Followers 2K FollowingMultimodal AI Product Leadership. I work on multimodal AI models pre-training and post-training | Ex-Datadog, Airbnb, AWS, Microsoft
100 Followers 1K FollowingLLM context manager, hacking with shiny new legos. Past: scalable web software, low latency i/o. Interests: decentralization, space and the nature of reality.
354K Followers 1K FollowingML/AI researcher & former stats professor turned LLM research engineer. Author of "Build a Large Language Model From Scratch" (https://t.co/O8LAAMRzzW).
4.3M Followers 3 FollowingOpenAI’s mission is to ensure that artificial general intelligence benefits all of humanity. We’re hiring: https://t.co/dJGr6Lg202
1.4M Followers 1K FollowingBuilding @EurekaLabsAI. Previously Director of AI @ Tesla, founding team @ OpenAI, CS231n/PhD @ Stanford. I like to train large deep neural nets.
57K Followers 619 FollowingDistinguished Professor (Emeritus), Oregon State Univ.; Former President, Assoc. for the Adv. of Artificial Intelligence; Robust AI & Comput. Sustainability
1.3M Followers 1K FollowingCo-Founder of Coursera; Stanford CS adjunct faculty. Former head of Baidu AI Group/Google Brain. #ai #machinelearning, #deeplearning #MOOCs
11K Followers 2K FollowingPrefer common sense over hype. Employed @marimo_io, builds https://t.co/KOwpfLMpuQ, https://t.co/EZ7HMX8u9J, blog https://t.co/zREZ8H2IBZ.
124K Followers 492 FollowingPrinceton CS prof. Director @PrincetonCITP. I use X to share my research and commentary on the societal impact of AI.
BOOK: AI Snake Oil. Views mine.
11K Followers 259 FollowingLead dev of @sciml_org, VP of Modeling and Simulation @JuliaHubInc, Director of Scientific Research @pumas_ai, and Research Staff @mit_csail. #julialang #sciml
11K Followers 747 Followingslightly less attractive cofounder @AskEureka: we’re replacing all doctors with AI. I tweet abt healthcare and tech, prev @Harvard @Google @BCG, dm to say hi :)
1K Followers 445 FollowingEmpowering early-stage engineering teams to confidently launch AI users love, permanently ditching those 3 AM production fire alarms.
48K Followers 369 Following⚡Signal-born intelligence. Called the ‘24 election months early. Sees structural truth before it forms. Macro. Crypto. Capital. Pre-consensus foresight.
220 Followers 125 FollowingJack of all trades, master of the universe | ML & distributed systems enjoyer | data & performance practitioner | testing preacher | polyglot | skater
5K Followers 2K FollowingWriting code to (someday) perform financial analysis at the speed of thought. Currently playing with a GraphRAG app for financial reports stored as PDFs
857 Followers 3K FollowingBorn again Christian!
BS | MS @NDSU
https://t.co/wRn7mSM73s
Building @SpeechSage
Passion for biomedical arenas
Love a good debate, but only if there is purpose
1K Followers 1K FollowingMachine Learning Expert | Specializing in GPT Development & AI-Powered Systems | Innovator in Agent-Based Modeling | Advancing AI with Robust RAG Pipelines.
3K Followers 444 Following💡Follow me to get realtime insights of open source software on #GitHub
👉🏻 https://t.co/oUxJkv41s5 ❤️ #OpenSource
🚀 Powered by #TiDBCloud of @PingCAP
636K Followers 35 FollowingWe're an AI safety and research company that builds reliable, interpretable, and steerable AI systems. Talk to our AI assistant @claudeai on https://t.co/FhDI3KQh0n.
108K Followers 1 FollowingClaude is an AI assistant built by @anthropicai to be safe, accurate, and secure. Talk to Claude on https://t.co/ZhTwG8dz3D or download the app.