Reposted by RibhuNaomi Saphra @nsaphra.bsky.social · 23/03/2026@gershbrain.bsky.social's new piece for The Transmitter includes a comment from me about something I think we often underrate in science: the actual degree of human understanding provided by a scientific model. 0276
Reposted by RibhuJoshua Foust 🪖🎮 @joshuafoust.com · 17/01/2026The “Jupiter Greedy” discourse is quite literally sending me into outer space 26705183
Ribhu @ribhulahiri.com · 10/12/2025Introduce yourself with: One Book 📚 One Movie 🎥 One Album 💿 One TV Show 📺 000
Ribhu @ribhulahiri.com · 09/11/2025Every NBA season I learn of a new way to spell the name "Jaylen" 000
Ribhu @ribhulahiri.com · 18/10/2025While these are just 3 thoughts I had while reading the above blog, what do y'all think about the "simple parts connected by clean interfaces" bit in the context of LLM-based systems today? What are some patterns and anti-patterns you have noticed? 000
Ribhu @ribhulahiri.com · 18/10/2025Unlike deterministic APIs, LLMs can return valid JSON that's semantically wrong. Strict schemas catch this at the interface, not in your business logic. 100
Ribhu @ribhulahiri.com · 18/10/2025More specifically the structure of it. What I mean is that when you start architecting an app, one of the first tasks we do is to create API contracts. Creation of a schema to communicate between components is what allows things to independently grow without any fear of breakdown of operability. 100
Ribhu @ribhulahiri.com · 18/10/2025Coming back to the thread, what makes any full stack application work well, even with hundreds of components, modules, microservices, and so on, is the reliability of the information. 100
Ribhu @ribhulahiri.com · 18/10/2025They operate at the logit layer of any open-weights model to ensure that the specified schema is "almost" deterministically followed. 100
Ribhu @ribhulahiri.com · 18/10/20253️⃣ Lack of reliability in outputs. Which is kind of the point Remy is making because that's what @dottxtai.bsky.social does so well. I would urge folks to try out the outlines library if you haven't already. 100
Ribhu @ribhulahiri.com · 18/10/2025Adding simpler flows from the beginning saves us this back-and-forth of adding validation checks and output parsers 100
Ribhu @ribhulahiri.com · 18/10/2025A single complex prompt that needs 3 retries costs more than 3 simple prompts that work first time. 100
Ribhu @ribhulahiri.com · 18/10/2025Unfortunately, what follows later on is having to add flows to retry, or post-process the output to refine it in a way where we get the output in the desired form and fidelity. Which leads us to more prompts anyway. 100
Ribhu @ribhulahiri.com · 18/10/20252️⃣ More prompts = More tokens = More cost. This is the mind-killer. This mental model is what leads engineers and products people to fit everything into less prompts. 100
Ribhu @ribhulahiri.com · 18/10/2025These are your initial set of sub-tasks, which you can later refine as needed. 100
Ribhu @ribhulahiri.com · 18/10/2025Now, you go about your task the way you – the master – would, and note down all the steps you needed to get it done. The first thought would be to combine a few of them. DON'T 100
Ribhu @ribhulahiri.com · 18/10/2025A mental model that I find helpful with this is the master-apprentice model. Here, you are the master and the LLM (you pretending to be one) is the apprentice. 100
Ribhu @ribhulahiri.com · 18/10/2025The ONLY way to get better at this is to train this muscle of breaking down tasks into the absolute singular task that is simple and stateless 100
Ribhu @ribhulahiri.com · 18/10/2025The reason why engineering and product managers exist. When given a goal, a lot of us sub-optimally break it down, based on our cognitive affordances. 100
Ribhu @ribhulahiri.com · 18/10/20251️⃣ Breaking down a task into simpler sub-tasks. The hardest of the 3. 100
Ribhu @ribhulahiri.com · 18/10/2025And this simplification of work is by no means easy. There are 3 main factors that make it harder: 100
Ribhu @ribhulahiri.com · 18/10/2025Agentic patterns that can alleviate this, can sometimes be worse, where an agent can have instructions on pursuing multiple things at the same time. 100
Ribhu @ribhulahiri.com · 18/10/2025There is almost a sort of pride that prompting folks take in being able to do everything in "one-shot" and not having to rely on multiple turns. 100
Ribhu @ribhulahiri.com · 18/10/2025Software engineering was built on this principle, but it's something I don't often see AI engineers follow (myself included). Many times there is a tendency to add every single instruction and decision point in a single prompt. 100
Ribhu @ribhulahiri.com · 18/10/2025I was prompted to structure these thoughts after reading this blog by @remilouf.bsky.social: blog.dottxt.ai/do-one-thing...blog.dottxt.aiDo One Thing Well 100
Ribhu @ribhulahiri.com · 18/10/2025"Complex systems should emerge from simple parts connected by clean interfaces" The principle based on which Unix was founded, and which guides building any software systems with a degree of complexity. Can this be replicated in AI systems? Here's some thoughts I had on the same 👇 120
Ribhu @ribhulahiri.com · 21/09/2025Shouldn't the number of donuts be proportional to the calories burnt? 000
Ribhu @ribhulahiri.com · 03/09/2025I was just trying to do a simple analysis on a piece of text and Qwen spent 15 whole minutes overthinking whether it should ask me for missing context or not 100
Ribhu @ribhulahiri.com · 03/09/2025Why can't these <7B reasoning models stop yapping to themselves? 100
Reposted by RibhuAlexander Doria @dorialexander.bsky.social · 24/08/2025Blogpost to read today: strong argument that excessive focus on the first tokens is not something learned from data distribution (like model should naturally "care" about the start of the text to grasp the rest) but a fundamental feature of attention graph. publish.obsidian.md/the-tensor-t... 2216
Ribhu @ribhulahiri.com · 22/08/2025What about the ULMFit paper? Since it laid the foundation for BERT 220
Ribhu @ribhulahiri.com · 06/08/2025That's fair, but the wheels and the bird's body generated by recent models are suspiciously better. Although I guess we don't know how much of that is improved capability vs conversations around your test sneaking into the training data. 000
Ribhu @ribhulahiri.com · 06/08/2025I think you may need a new test. Feels like every model after your AIEngineers World Fair talk has been doing a lot better with the pelicans riding a bike SVG task. 100
Ribhu @ribhulahiri.com · 02/08/2025was doing some interesting work around something similar. This just catalyses it 🚀 More soon 👀 020
Reposted by RibhuEthan Mollick @emollick.bsky.social · 20/07/2025Don't leave AI to the STEM folks. They are often far worse at getting AI to do stuff than those with a liberal arts or social science bent. LLMs are built from the vast corpus human expression, and knowing the history & obscure corners of human works lets you do far more with AI & get its limits. 413827
Ribhu @ribhulahiri.com · 16/07/2025Miss people don't realize it, but this is basically R1 all over again 010
Reposted by RibhuBen Recht @beenwrekt.bsky.social · 10/07/2025Fully open machine learning requires not only GPU access but a community commitment to openness. (Some nostalgic lessons from the ImageNet decade.)argmin.netAn open mindsetThe commitments required for fully open source machine learning 1264
