Harry Jackson

AI & Programming & Founder KDSoftware

[Harry Jackson @keydol123]

GPT-6 Astra vs Claude Opus 5.5

Same prompt: "Build an interactive keyboard made of honey."

GPT-6 made something that looks more like jelly

Opus actually nailed the honey look

Astra: $10.67
Opus: $12.48
Which one wins?

[Harry Jackson @keydol123]

@cocco_ai Haha, this is hilarious

[Harry Jackson @keydol123]

@olezhavich Both look really beautiful! The lighting and colors are amazing.

[Harry Jackson @keydol123]

@ka1manov Turning $170 into $2,154 overnight sounds like every trader’s dream

[Harry Jackson @keydol123]

@0xLilByte Astra definitely earned that medal 😂

[Harry Jackson @keydol123]

I asked GPT Astra 6 and Claude Opus 5.5 to create a bottle filled with honey

Looks like Opus has no idea what liquid honey is supposed to look like

[Harry Jackson @keydol123]

@KostyaAI Yeah, the viscosity still needs some work

[Harry Jackson @keydol123]

@theaidesk_ Agreed! I'm thinking of trying something even more challenging next time

[Harry Jackson @keydol123]

@PMsaysIT Glad you liked it!

[Harry Jackson @keydol123]

@ka1manov Yeah, Astra really nailed the liquid physics this time

[Harry Jackson @keydol123]

@dromexa Same! Especially how the honey actually flows out when you cut the bottle

[Harry Jackson @keydol123]

@0chob Haha, I wasn't expecting such different results from the same prompt

[Harry Jackson @keydol123]

@ka1manov I wonder how much time six agents spend reviewing a PR compared to a regular code review. Would be interesting to see some real benchmarks.

[Harry Jackson @keydol123]

@Nitocnt Yeah, especially getting the viscosity right

[Harry Jackson @keydol123]

@cocco_ai Coming right up!

[Harry Jackson @keydol123]

@olezhavich Exactly

Opus somehow decided the bottle should be made of honey too

[Harry Jackson @keydol123]

@0xRhinoX Gonna see if I can land that barrel roll

[Harry Jackson @keydol123]

@ka1manov This could be really useful for tracking competitor prices. I'd be curious how well it handles websites that change their layout frequently

[Harry Jackson @keydol123]

@velesxbt Would be interesting to see how this performs in a live market with real latency, fees, and competition from other market makers

[Harry Jackson @keydol123]

@0xLilByte I'd put the general rule in CLAUDE.md and let the agent choose which tests to run based on the changes. But I'd still run the full test suite before merging

[Harry Jackson @keydol123]

@dromexa Flappy Bird with a flying Tesla and rockets is a whole new level of chaos 😂 Gonna try beating 51