:)

Founding DevRel @scaledown_team

[:) @purvahk]

the founder who builds a solution for every problem he runs into 🫡

[Parth Koshti @itsParthVader]

I googled "sitemap explorer" and hated every result.

Most have clunky UX. No easy way to see site structure and Open Graph previews together.

So I took an hour to build what I wanted.

Explore any sitemap visually. Inspect OG metadata side by side.

[Parth Koshti @itsParthVader]

nobody's filling out ur demo form on Hacker News, but they'll tell u exactly why they hate ur current vendor. here's where to find leads like that 👉 https://t.co/K9AdoQP1jD

[:) @purvahk]

@RoundtableSpace the only thing you need for tokenminning - https://t.co/GTBqwIJKZ5

[:) @purvahk]

@kalepail get on the dietcode waitlist, it's gonna be worth the wait

[:) @purvahk]

@csoham_ true that

[:) @purvahk]

@neal_k_patel @csoham_ the man who speaks only in PRs is building in public🔥

[:) @purvahk]

@csoham_ tokenminning szn🫡

[:) @purvahk]

if your claude code bill annoys you, this is for you🥤

[Soham Chatterjee @csoham_]

day 1 of building DietCode🥤

i'll be documenting the journey here, shipping every day, learning from users, and making Claude Code dramatically cheaper without compromising quality

waitlist → https://t.co/43vtTfmHnk 🚀

[:) @purvahk]

@Itsbrutox @meetgranola the best one

[ScaleDown @scaledown_team]

🚀Our cheapest tokens just dropped another 50%, meet the @scaledown_team Batch API

• Run extract, summarize & classify at scale in a single batch
• Structured results back, every entity scored and classification explained

For when you've got a backlog. Link in comments👇

[Neal Patel @neal_k_patel]

Meet the Grok killer

@scaledown_team's task-specific SLMs beat Grok (xAI's frontier model with the biggest GPU cluster on the planet):

Accuracy: 7.93% higher
Cost: 24x cheaper
Latency: 2.2x faster

Full benchmarks: https://t.co/RniWQwVMYK

[:) @purvahk]

@dazxlr building kode https://t.co/VLWxyXjjvW would love your thoughts!

[:) @purvahk]

@appwrite building kode!!
https://t.co/VLWxyXjjvW

[:) @purvahk]

@aaliya_va @heynavtoor true! you should give kode a try https://t.co/VLWxyXjjvW

[:) @purvahk]

@midudev how does it compare to kode?
https://t.co/VLWxyXjjvW

[:) @purvahk]

@ushercakes @ushercakes curious about your thoughts on kode?
https://t.co/VLWxyXjjvW

[Neal Patel @neal_k_patel]

Introducing ScaleDown

15x cheaper. 63x faster. 5.1% more accurate than GPT-5.4 Mini.

Task-specific SLMs for the 70-80% of AI workloads that don't need a frontier model.

From NeurIPS '25 to https://t.co/imZLOYCcTA

[ScaleDown @scaledown_team]

General models do everything. That's the problem.

ScaleDown builds task-specific models for the steps in your pipeline where general models fail most.

/compress is already live. Two more are dropping soon.
Try it here: https://t.co/k36oXKgshG🚀