Something interesting about Gemini 3.1 Pro 👀
This chart shows Gemini 3.1 Pro Preview scoring 32 on hallucination resistance, while Gemini 3.7 Flash scores 26.
So in some cases, the “older” Pro model might actually solve problems more accurately than the newer Flash.
Has anyone else noticed this?
Something interesting about Gemini 3.1 Pro 👀
This chart shows Gemini 3.1 Pro Preview scoring 32 on hallucination resistance, while Gemini 3.7 Flash scores 26.
So in some cases, the “older” Pro model might actually solve problems more accurately than the newer Flash.
Has anyone else noticed this?
Not enough people know about this. 🧐
OpenCode Go lets you pay $10/month and get up to $60 worth of model usage.
That's basically getting $6 of usage for every $1 you spend.
If you do a lot of AI coding, this is actually insane.
Why is nobody talking about this?
Not enough people know about this. 🧐
OpenCode Go lets you pay $10/month and get up to $60 worth of model usage.
That's basically getting $6 of usage for every $1 you spend.
If you do a lot of AI coding, this is actually insane.
Why is nobody talking about this?
Gemini Omni 1.1 Flash has been released!
- It secured 2nd place on https://t.co/TVDal2RFmc with a score of 1488.
- Video generation is possible with 360p and 720p options.
- Videos up to 10 seconds long can be generated, though an extension option is also available.
Do you think Gemini Omni Flash is worth using?
Here is the https://t.co/TVDal2RFmc ranking list👇👇👇
Gemini Omni 1.1 Flash has been released!
- It secured 2nd place on https://t.co/TVDal2RFmc with a score of 1488.
- Video generation is possible with 360p and 720p options.
- Videos up to 10 seconds long can be generated, though an extension option is also available.
Do you think Gemini Omni Flash is worth using?
Here is the https://t.co/TVDal2RFmc ranking list👇👇👇
The builders winning right now aren't just fast coders—they are relentless iterators who treat failed experiments as cheap data points.
Keep shipping, measure retention ruthlessly, and double down on what actually sticks.
For those building tools or indie projects: What’s your biggest bottleneck right now—building, distribution, or retention?
Building software in the AI era created a massive illusion: shipping code got 10x faster, but building something people actually retain got 10x harder.
After launching multiple web tools and experiments, here are 4 painful engineering & product truths most builders learn the hard way 🧵👇
3/ Smart architecture beats brute-force API calls.
Throwing raw LLM API calls at every feature burns runway fast.
- Sustainable builds rely on hybrid setups:
- Rule-based logic & regex for deterministic tasks
- Fast, lightweight models for intent routing
Heavy frontier reasoning strictly reserved for edge-case fallback