Joshua Martin

Sansa is officially live 🎉🎉 an AI API for frontier performance at half the cost. anywhere you use AI you can use Sansa instead. setup takes seconds for the openai sdk: - swap your base URL - add your Sansa key Use Cases: 🦞OpenClaw ⚙️automations 🤖agents 🧠anything AI sign up: app.sansaml.com if you were on the waitlist, DM me for free credits. demo attached. go build something!

@JanarJurisson Nice! If you were into the nitty gritty of model selection do you think it would be less useful?

@Jbm_dev Yeah, I like it. I'm not into the nitty-gritty of models (yet) and tokens, so I like it this way.

I was looking for credits based, "connect it and forget it", kind of service. And @Jbm_dev's Sansa for me is exactly that. No problems this far, but my AI model selection awareness for specific jobs, are also below the bar probably. I used inference. net for a while, but they were adding and removing models and my requests broke since the model was required parameter if I remember correctly. Shoutouts thrive this community 🙌

@JanarJurisson Same here man, I have just gotten so busy. I used to respond to every comment I got, now I can barely keep up.

I'm a consistency guy here on X, but the last month or so, I’ve been more of a vibe poster.

@DimitriGilbert @Zai_org What IDE are you using it in?

Aaaaand GLM is completely dumb again... @Zai_org seriously, guys, this is getting old at some point... I'm rooting for you and I want you to succeed, but the coding plan gets unusable too often ! It lasted 2 days last week and it's the same right now... non sense and/or non compliance past 30k tokens... I prefer a slow model to a quantized brainless version

@mdnlabs 1/5 mostly because I don’t like bacon

Rate my bacon

@WesamMikhail at least it’s fun?

My college professor used to say: "Sales, marketing, and distribution are the hard parts of computer science." Little did I know at the time how right he was!

@Jchammond_ Switch back to Cursor and use composer 2

Yo what are we going to do when this 2x rate limit ends in the Codex app?

@JanarJurisson What is the project for? Still stealth?

My luck has run out. Flatlined like my stealth pSEO content/database project's indexing progress. 📜 Total URLs in sitemap: 617 (+175) 🔖 Indexed pages: 364 (+1) 👀 Impressions: 167 avg/day last 7 days. 🎯 Clicks from Google: 0.42 last 7 days 😬 One thing is sure. I took a bit big bite and it's all based on hunch. But I am working to automate it ⬇️

Fair enough! Not sure how strict your data governance/rules are, but we do have an option to not save any traces/requests on our server. Giving users complete control over their data is a top concern for us, so if there is anything that would move the needle for you on this let me know and we will get it implemented!

@Jbm_dev @JanarJurisson @ScrapingF Hmmm ok, that would be quite a significant save... just can't afford to send the data I'm sending through another proxy :(

@FriedrichFrbel For sure, What IDE do you use right now? @AlexWinkyFace and I were just talking yesterday about making this happen!

@Jbm_dev That sounds interesting! Is there a simple way one can add this to their IDE?

It’s crazy how fast new models come out. In the last 3 weeks we have had: Sonnet 4.6 Gemini Pro 3.1 Gemini 3.1 flash lite Kimi k2.5 Qwen 3.5 (a bunch of them) Nvidia Nemotron 3 Super (120b) Hunter Alpha GLM 5 (Sunday) ByteDance Seed 2 GPT 5.3 chat GPT 5.4 GPT 5.4 mini & nano (yesterday) MiniMax M2.7 (today) Mistral 4 Small (Monday) All this intelligence, all this change. Sansa is the best way for companies to keep up, hands down. It’s like your apps AI being on automatic updates

@nikitabier The em dash…

The fortress we are building—and the layers of redundancy—to protect the platform against the AI Slopacalypse will seem obvious in a few months. Whether we use every tool in our toolkit is TBD, but it would be negligent to not have them ready.

@MythThrazz @JanarJurisson @ScrapingF About $180, that’s if it’s an even-ish input output split, but a lot less if it’s heavy on the input. That said, they wouldn’t all be processed by Opus, but the quality would match

@Jbm_dev @JanarJurisson @ScrapingF How much would I pay for 38M Opus 4.6 tokens? :D

We have a routing model that reads the input (messages, system prompt, tools) and predicts what model can handle the task. We trained the model on more than 15M requests (20+ billion tokens) of real interactions, where each interaction was evaluated against a ground truth outcome. It’s very good at the task! For example, using Sansa we hit nearly 98% on MMLU pro, which is better than Gemini 3.1 Pro, but the cost is less than half.

@Jbm_dev How does sansa determine which model to choose?

do you trust AI to architect your system? where's your line?

@joncphillips Yes! 100% it’s my vacation haha

@Jbm_dev Hah might be a bit of both, but then again, what’s a vacation for if not for doing exactly what you feel like doing!

on vacation right now woke up before sunrise walked 2 miles to a cafe in a town i don't know just to get a few hours of building in is this passion or a problem?

@JanarJurisson Yeah, there is no desk where I’m staying! Haha

@Jbm_dev If you wanted to walk & have local cafe experience, it sounds great. If only place where can work is cafe, it's a problem 😅

@JanarJurisson @ScrapingF How’s your experience been so far? Any issues send me a DM :) and thanks for the shout out!

I already implemented scraping source urls with @ScrapingF and processing it with AI (thanks @Jbm_dev for trysansa.com) to normalise specs and create entries to my database. 🙌 Because eventually, it must be semi-automated anyway. To check spec changes on manufacturer homepage. Still manually checking the results. But most of the work is done in background.

@markccalugay Yes indeed!

@Jbm_dev Those small moments are what really keep up the grind worth it!

@jonathanbrnd Yeah I think so too! Haha

@Jbm_dev same here. impossible to disconnect! i guess it's passion or obsession haha

@buildwithjaden Thanks Jade! I think so too!

@Jbm_dev If you love what you do, it's passion