๐ Hello everyone, I'm thrilled to share some exciting news with you all! I recently graduated with my Masters in Cybersecurity and Information Assurance from @๐๐ด๐, and I'm beyond thrilled about this achievement! ๐
๐งต ๐ Thread ๐งต ๐
@Zai_org I guess I need to buy a new laptop, 32GB RAM is not suffice for all the electron apps, npm packages, IDEs, orchestrators, design tools, especially browsers :)
I hand drawn this design and converted to SVG for https://t.co/LczZ6eXahe - it has a subtle animation as well, please try. I built this for @awscloud and @vercel hackathon. ๐ค
Fable is back; start your engines to measure Fable performance - only at https://t.co/LczZ6eXahe
๐ Anthropic claude-fable-5 ๐ 101.9 tok/s, 3606ms TTFT on https://t.co/LczZ6eXahe
https://t.co/yKM6WU4jLV
@TTrimoreau https://t.co/LczZ6eXahe has been significantly improved and getting lot of traction.
๐๏ธ Race Mode
๐ Submit your LLM perf to the leaderboard
๐ Added support for Grok, Groq, Z, Google, Cerebras, Mistral and more
โฆ๏ธ Share your speed results with 1 click
We're close to hitting 10K on X. Should celebrate.
Giving away 3 Pro subs
STEPS
1. Follow @CommandCodeAI
2. Repost this post
3. Reply with what you're building with Command Code
We'll pick randomly soon as we hit 10K. LFG!
https://t.co/LczZ6eXahe has been significantly improved and getting lot of traction.
๐๏ธ Race Mode
๐ Submit your LLM perf to the leaderboard
๐ Added support for Grok, Groq, Z, Google, Cerebras, Mistral and more
โฆ๏ธ Share your speed results with 1 click
Most AI coding agents burn frontier model tokens on tasks a smaller model could handle.
@cognition_ai just fixed that with Devin Fusion.
Here's the architecture that makes it work (and why it's clever from a systems design POV):
/thread
6/ The analogy from the blog:
"You wouldn't drive a Lamborghini to the grocery store."
Why use a model that finds zero-day vulnerabilities to rename a CSS class?
The age of one model for all work is ending.
Try it: https://t.co/JPx4JDjxGU
Full breakdown: https://t.co/oIyDFMZRU6
Most AI coding agents burn frontier model tokens on tasks a smaller model could handle.
@cognition_ai just fixed that with Devin Fusion.
Here's the architecture that makes it work (and why it's clever from a systems design POV):
/thread
5/ The numbers:
35% lower cost vs GPT-5.5 and Opus 4.8 on FrontierCode benchmark.
88% of merged PRs at Cognition driven by the Fusion router.
As someone who measures LLM performance at https://t.co/LczZ6eXahe, "cost per task" is the metric I care about most.