Eran Sandler

@erans

Builder, operator and investor. Infra, AI, and product nerd. Opinions are my own. Checkout @lunaroute_ai (lunaroute.com) to stop caring about token limits

eran.sandler.co.il

Joined April 2007

4,368 Following

1,009 Followers

Replies Reposts Media

Eran Sandler
@erans
Oct 12
If you are a @ycombinator company and are looking for private, fixed cost inference of open weight models (specifically GLM 5.3, GLM 5.3 Flash and DeepSeek 4.1 Flash) - DM me. I'll hook you up with a trial account on @lunaroute_ai (lunaroute.com)
1
Eran Sandler
@erans
Oct 12
When you are not token scarce you can think about things differently. Like a data sovereign replacement for chatgpt chats that encrypts each chat with its own key. This is what I ended up building for myself at @lunaroute_ai - Luna Chat :) If you want early access DM me or
Eran Sandler
@erans
Oct 8
1/ Agents in 2026 are being built like systems software: Rust harnesses, one VM per user, cron jobs, shell scripts. Here's the short version of how we got here. 🧵
1
Eran Sandler
@erans
Oct 6
This is something I've been thinking about a lot lately. You really don't know how a model will behave for your workload until you try it. That's also why I think having the freedom to experiment matters so much. This is the longer version of an @ignitetalks I gave at the AI
LunaRoute
@lunaroute_ai
Oct 6
Same model. Same hardware. We changed how we scheduled the work and latency went from 2.7s → 0.4s. There’s a lot more to inference than the model name. The model is not the product. lunaroute.com/blog/the-model…
Eran Sandler
@erans
Oct 6
I'm really excited for Beam. Don't follow the benchmarks as this is the first model. Muse Spark 1.0 was also meh, but fast iterations into 1.3 and its quite good (1.2 was also good). If @reflection_ai keep their pace it would be awesome soon. Also, make sure to check it on
Reflection
@reflection_ai
Oct 5
Introducing Beam: a highly efficient agentic open model with 501B total parameters and 23B active. - Frontier reasoning efficiency - Advances the Western open frontier on coding & agentic tasks - Trained end-to-end from scratch Full weights release this month. Learn more about
3