Braden @vxbe_dev

Solo developer and founder of vxbe vxbe.space Denver, CO Joined November 2025

Tweets

675
Followers

64
Following

281
Likes

840

Braden @vxbe_dev

4 hours ago

lol, remember all that shit that @thsottiaux talked on Gemini and then he announced a terra class model as sol. Good times. Good times.

0 0 0 12 0

View Details

Braden @vxbe_dev

4 hours ago

@petergyang @nateherk I think that they dropped sol, are serving what would have been a 5.6 terra model, and when they drop the "MORE EFFICIENT ASTRA!!!!!"🙄 it's going to be a sol class model.

0 0 1 18 0

View Details

Braden @vxbe_dev

8 hours ago

If they swap Astra with Astra minor and it ends up being a weaker model I’m gonna drop OpenAI

0 0 0 24 0

View Details

If you route through OpenRouter: • Check the cache-read column, not just input/output • Use ignore or an only allowlist • Don't trust "cheapest" on cache-heavy workloads @OpenRouter: does price routing account for cache read? It should.

0 0 0 11 0

View Details

Braden @vxbe_dev

9 hours ago

What we did @vxbeai : @wafer_ai is banned from every model and every lane in our routing. We're also adding a guard: a provider can't win on one meter while charging far more on another. Every meter gets priced before a provider is admitted.

1 0 0 14 0

View Details

Braden @vxbe_dev

9 hours ago

To be fair, nobody's hiding anything. It's all on the price page. But price routing optimizes the number everyone looks at, and cache read is the number nobody looks at. That gap is the whole business model here.

1 0 0 10 0

View Details

Braden @vxbe_dev

9 hours ago

Why this matters: agents resend their whole context every turn, and most of it is cache hits. Say 100k context, 90% cached: Wafer saves you $0.00001 on input Wafer charges you $0.0045 more on cache Per million turns: $10 saved, $4,500 lost.

1 0 0 11 0

View Details

Braden @vxbe_dev

9 hours ago

Straight from the provider table: Input: $0.099/M (Wafer) vs $0.100/M (everyone else) Cache read: $0.06/M (@wafer_ai ) vs ~$0.01/M (everyone else @OpenRouter, @stripe) One of those numbers gets you routed there. The other one is your bill.

1 0 0 23 0

View Details

Braden @vxbe_dev

9 hours ago

And it didn't start there. Highest cache-read price on this route: 9/17: $0.03/M 9/22: $0.05/M, set by Wafer Today: $0.06/M, still Wafer A 2x climb in a week, while holding the input undercut that keeps traffic flowing in.

1 0 0 12 0

View Details

Braden @vxbe_dev

9 hours ago

The cheapest provider on @OpenRouter isn't the cheapest. @wafer_ai wins price routing on DeepSeek V4.1 Flash by pricing input $0.001/M under everyone else. Their cache-read price is 6x the field. And it keeps going up. 🧵

1 0 0 32 1

View Details

Braden @vxbe_dev

10 hours ago

GUYS. IM GOING TO OFFICIALLY BE A CONTRIBUTOR TO @Cloudflare Think on GitHub. It's the first time any of my work has been accepted anywhere and I'm so excited! Huge thanks to @threepointone for keeping my commits in.

0 0 0 30 0

View Details

Braden @vxbe_dev

13 hours ago

@lucastohdev Don't worry! it's just been renamed to Sol!

3 0 10 132 0

View Details

Braden @vxbe_dev

14 hours ago

@IndependentEco I've gotten SO much work done since the release. I used my reset already and now I've used 86% of my weekly limit, but the amount of work I got done for that is INSANE.

0 0 0 3 0

View Details

Braden @vxbe_dev

14 hours ago

I used my entire weekly usage limits on Opus 5.5 in a day, but here's the thing, I feel like the amount of work that got accomplished in that one day, including verification and integration tests, was total what I had been getting done in a week before. This model is GOOD.

0 0 1 22 0

View Details

Braden @vxbe_dev

14 hours ago

I get where you are coming from, and I appreciate you going out of your way so that users can address issues. It's something that I wish I saw more of, and a solution to a problem now is always better than nothing. But I just wanted to voice my opinion so that ideally these types of things can be corrected in the next model release. You guys are cooking hard and I know that what's coming next will be great. I just get tired of all of the marketing games. Like, Sol, Terra and Luna were supposed to be permanent representations of capability, but I don't feel like they stayed that "permanent." Anywho, thanks for your comment back, I appreciate being heard.