NaveenKumar Namachivayam

full stack developer @salesforce @mulesoft | Tweets are my own | Building | Feather Wand - JMeter AI Agent | | |

[NaveenKumar Namachivayam @QAInsights]

👋 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 @𝘄𝗴𝘂, and I'm beyond thrilled about this achievement! 🎉
🧵 👇 Thread 🧵 👇

[NaveenKumar Namachivayam @QAInsights]

@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 :)

[NaveenKumar Namachivayam @QAInsights]

When you get this kind of message in your CLI? Keep this setting enabled to save compute/tokens.

[NaveenKumar Namachivayam @QAInsights]

@stephenhaney adding it to https://t.co/e9umc2Zr8X :)

[NaveenKumar Namachivayam @QAInsights]

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. 🤞

[NaveenKumar Namachivayam @QAInsights]

@lightsilver323 https://t.co/e9umc2Zr8X

[NaveenKumar Namachivayam @QAInsights]

Fable is back; start your engines to measure Fable performance - only at https://t.co/LczZ6eXahe

🚀 Anthropic claude-fable-5 📊 101.9 tok/s, 3606ms TTFT on https://t.co/LczZ6eXahe

https://t.co/yKM6WU4jLV

[NaveenKumar Namachivayam @QAInsights]

@JinjingLiang @orca_build Work comes first; X later :)

[NaveenKumar Namachivayam @QAInsights]

@TTrimoreau https://t.co/LczZ6eXahe has been significantly improved and getting lot of traction.

🏎️ Race Mode
🏁 Submit your LLM perf to the leaderboard
🚘 Added support for Grok, Groq, Z, Google, Cerebras, Mistral and more
♦️ Share your speed results with 1 click

[NaveenKumar Namachivayam @QAInsights]

@CommandCodeAI Building https://t.co/LczZ6eXahe

[Command Code @CommandCodeAI]

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!

[NaveenKumar Namachivayam @QAInsights]

https://t.co/LczZ6eXahe has been significantly improved and getting lot of traction.

🏎️ Race Mode
🏁 Submit your LLM perf to the leaderboard
🚘 Added support for Grok, Groq, Z, Google, Cerebras, Mistral and more
♦️ Share your speed results with 1 click

[NaveenKumar Namachivayam @QAInsights]

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

[NaveenKumar Namachivayam @QAInsights]

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

[NaveenKumar Namachivayam @QAInsights]

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

[NaveenKumar Namachivayam @QAInsights]

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.

[NaveenKumar Namachivayam @QAInsights]

@Joi2James https://t.co/LczZ6eXahe