Opus 5 Ultracode vs 5.6 Terra High (not Ultra) on code review, security audit, UI/UX evaluations...
Terra kept pointing out mistakes that Opus made, and Opus agrees. Have not seen the same corrections the other way. This is quite surprising especially given the effort settings.
148
Have been using GPT-5.6 Sol and Terra for the past couple of days - for some tasks, 5.5 was smarter and completed them instead of sitting there waiting for me to tell it what to do next.
5.6 does not feel smarter in coding my apps. Perhaps my expectations were high. @OpenAIDevs
191
Schrödinger's Cat goes to the Vet
Turn up the volume & dance with the magic cat!
Suno lyrics and style prompt:
suno.com/s/v3gtpudK5OO98bV3
#synthpop #dance #suno
103
My standalone password generator:
Plain HTML, CSS & Javascript, does not rely on any library or external APIs.
Features include: Password length, upper/lower case, numbers & special characters (customizable), strength, and copy to clipboard.
daveyang.github.io/pswgen/
#password
1
129
Claude Sonnet 4 refactored a project with 12 modules and optimized the codebase, resulting in 5 modules and a lot less code (that was generated by older LLMs).
It cost about $5 but the most amazing thing is it did not break anything on first shot!
Wow!
@AnthropicAI #Claude4
2
142
Oh how I love these colors! They give me much joy simply being colors that my eyes can see.
Perhaps one day, when my eyes are too old to appreciate them, I'll focus on the spirit flowing among these wonderful colors and the movement it brings.
Partial view of a bigger picture.
1
83
New shader inspired by @kishimisu, created in Love2D as part of the video effects app.
3
223
Some of the latest LLMs claim to score the highest in coding but actual tests show they fail miserably.
Google Gemini 2.5 Pro (Experimental 03-25) continues to make up non-existing API functions and is unable to solve the task after 2 days. No recent LLMs fail like it.
1
160
After testing Claude 3.7 Sonnet for about an hour on coding, it is really great and got things right in a single shot and fixed a bug that other LLMs struggled with (and were unable to fix). @AnthropicAI has done it again!
93
