ComputeSDK

A market and router for compute

[ComputeSDK @computesdk]

🏁 ComputeSDK Actions are live

What we built:
⚡ Massive concurrency for parallel dev cycles
🌐 Sandbox providers + regions
🌿 Git-first workflow portability
🛡️ Ordered fallback for reliability
🤖 Agent-native CLI + API

Let’s get into it ↓

[ComputeSDK @computesdk]

@tiliondev @usekernel @browserbase full results here: https://t.co/w9sPZ74VDi

[ComputeSDK @computesdk]

This week's run of Browser Throughput just finished 🏁

🥇@tiliondev
🥈@usekernel
🥉@browserbase

Top 3 of 7 providers

[Kush @kushbhuwalka]

taking control of a sandbox in an AI product is such a poor experience.

every click has visible latency. this screenshot is from @DevinAI.

same experience from @daytonaio in @puffleai.

our team has started testing UI locally because of this, which is really annoying.

have any sandbox people solved for this?

[ComputeSDK @computesdk]

@isorun @promigence @namespace see the full results here: https://t.co/ZLPm5MI2II

[ComputeSDK @computesdk]

This week's run of Sandbox Dax just finished 🏁

🥇@isorun 36.134s
🥈@promigence 36.190s
🥉@namespace 46.214s

[ComputeSDK @computesdk]

@github @neondatabase @OpenRouter see the full results here: https://t.co/VUjE90HWBY

[ComputeSDK @computesdk]

This week's run of AI Gateway Latency - Kimi just finished 🏁

🥇@github Copilot 92.7
🥈@neondatabase 86.2
🥉@openrouter 82.8

[ComputeSDK @computesdk]

full results: https://t.co/hkiianvzy1

[ComputeSDK @computesdk]

AI Gateway Latency Kimi has finished!

🥇@github
🥈@neondatabase
🥉@ngrokHQ

[ComputeSDK @computesdk]

Full results here: https://t.co/pYEKtAOj7A

[ComputeSDK @computesdk]

And now storage lifecycle

🥇@Azure
🥈@archildata
🥉@awscloud

[ComputeSDK @computesdk]

Sandbox TTI Burst just finished

🥇@Miosa_AI
🥈@isorun
🥉@promigence

[ComputeSDK @computesdk]

see the full results: https://t.co/rNon2YpbtY

[ComputeSDK @computesdk]

This weeks run of Browser Throughput just finished

🥇@tiliondev
🥈@browserbase
🥉@steeldotdev

[ComputeSDK @computesdk]

This is why we built our TTI is that we found some people were lying.

problem solved now. we publish every week!

[Shivansh Vij @confusedqubit]

If you can’t start a customer’s workload (cold) within your supposed sandboxes spin-up time, you’re straight up lying to your users.

Time to interactivity is the *only* number that matters.

[ComputeSDK @computesdk]

See the full results here: https://t.co/hkiianvzy1

[ComputeSDK @computesdk]

AI Gateway Latency - Kimi Benchmark

🥇@github
🥈@neondatabase
🥉blazerail

[ComputeSDK @computesdk]

See the full results: https://t.co/pYEKtAOQX8

[ComputeSDK @computesdk]

Storage Lifecycle Benchmark

🥇@archildata
🥈@Azure
🥉@awscloud

[ComputeSDK @computesdk]

See the full results here: https://t.co/mwIsTzzQ4H

[ComputeSDK @computesdk]

Sandbox Dax (@thdxr) Benchmark has complete.

🥇@isorun
🥈@Miosa_AI
🥉@blaxelAI