Ron Itelman

@pzakin Hi Peter, the intent management layer is missing right now for continuous intelligence. Minimizing risk in misaligned intent in real time with user uploaded sources outside of training, data sets and data management is key. Banking/finance tested. Looking for first check.

@JenniferHli - working on runtime data entropy calculation and minimization. Your Big Ideas 2026 piece resonated, and I’d love to connect.

Just released open source (MIT license) real-time word detection for governance, linting, semantic clarity, API calls before sending to an LLM, and quick menus. trustquery.com

Please connect with me at bsky.app/profile/ronite… as I will no longer be using X / Twitter. Thanks!

@garrytan For those of us with families and are unable to go to SF, will there be a remote cohort at some point?

Some of the best YC companies decided to apply the day before.

@mwseibel For those of us that have families, is a remote batch coming up? Or maybe you can open a cohort in Denver? :)

Time to apply to the next YC batch:

@jkbjournalist Thanks!

@ron_itelman @jkbjournalist.bsky.social

@martin_casado Hi @martin_casado - That's why we're working on intent query language, to discern probabilistic and deterministic relationships and actions. Would love to chat!

You can't fix hallucinations with model scale. Even reducing them by orders of magnitude still requires full error handling if correctness is required. No amount of GPUs or data will magically fix that. Like the Internet, we need to move up the stack for formal guarantees. There are eery parallels to the early Internet where the fiber folks where focused on building out capacity. And disconnected from the software folks how new that applications hadn't matured enough to take advantage of best effort. So, yes model progress may slow down because we're running out of data. But I suspect the real limitation is going to be at the app layer. It takes a long time for the industry to evolve its application stack.

@Kasparov63 I remember your many posts warning us of the dangers of "annihilating the ability to know what is true". At dinner recently, someone stated that as a country our "shared communal reality is gone", and I immediately thought of your warning.

Just found youtu.be/nL9jEy99Nh0?si… by @fchollet - an important leader in his field, worth your time to watch!

@fchollet Please start posting on blue sky! :) Trying to get off here, and your posts are gold!

I miss when this app could be used to share links to your work, find interesting people with shared interests, that sort of thing. I enjoy being on the receiving end of a nonstop unhinged political propaganda machine as much as the next guy, but this is a bit too much.

@martin_casado @martin_casado Your observation aligns well with what we're researching, an Intent Query Language (IQL) for AI interactions. Your analogy is helpful to consider new ways we might benchmark this as a "formal layer", borrowing from network programming. Thanks!

Dealing with the unpredictability of LLMs in code reminds me a lot of network programming. You end up having to implement retries, back off, and sometimes have to overlay transaction identification for idempotency. I wouldn't be surprised if this emerges as a formal layer.

I just signed up for bluesky, and not having unhealthy content I don't want to see constantly in my feed is quite refreshing. Will be moving over there, hope to see you there too!

@saranormous Hi Sarah, we pivoted to analyzing developer code as a way to source talent, we've got a pretty unique dataset and recommendation engine. Would love to help any of your companies with a quantitative way to hire. Our motto is "Code speaks louder than words". Resumes don't work. DM

some feedback for candidates looking for jobs at promising startups: don’t claim “x years of experience” that’s not an outcome, that’s hanging around the hoop. you can do something in a mediocre way for years instead describe something great you have built, fixed, quota hit…

“Go for the pain of discipline, it weighs ounces. The pain of regret weighs tons.” — Jim Rohn

You can have a great startup idea, but a great startup idea is like having a great plan for a fistfight. The moment the idea actually touches customers is like the moment you get punched in the face. What matters is what you do at that moment, not the plan.

A mindworm that has caused great damage among AI researchers is the implicit, universal assumption that every piece of data is a (random) "sample" from a (static) "distribution". This is a valid way of modeling some phenomena, but it isn't applicable to the vast majority of real-world signals.

Nighttime dad will tell daytime dad it's no problem playing video games at night, because he can "sleep in". This is most definitely a lie. DO NOT FALL FOR IT!