Kirkeelee

Previously reading genetic code to uncover diseases, now reading smart contract code to uncover vulnerabilities.

Portfolio:

[Kirkeelee @kirkeelee]

Went to bed staring at 4% weekly usage left on zcode,
woke up to 4× 5-hour resets and 4× weekly resets. Did not see that coming. Thanks @Zai_org

[Kirkeelee @kirkeelee]

Tokenmaxxing the current unlimited windows.
-Qwen3.8-Flash on Qoder: bad. Shortcuts tests to make them pass.
-Keepers: SWE-2 on Devin, GLM-5.3-Flash on ZCode.
-New stealth Space Bunny on OpenCode. Above expectations so far.

[Kirkeelee @kirkeelee]

After this, I can never look at those other AI labs the same way again.

[Alexandr Wang @alexandr_wang]

state of the art motherfuckers

[Kirkeelee @kirkeelee]

Me trying to understand why the new jev AI is different.

[Clash Report @clashreport]

Putin casts online vote in Russia’s State Duma elections.

[Kirkeelee @kirkeelee]

Fable 5.1 is not humble. Gave a list of latest models(Astra excluded) to use in a threat modeling phase of my workflow, it didn't like them:

[Kirkeelee @kirkeelee]

@DevDacian You can get unlimited SWE-2 (and GLM5.2) usage on any of the Devin plans, too. Good for long running, capability independent tasks. But I like GLM5.3-flash better.

[Kirkeelee @kirkeelee]

@boltdotnew Forge

[Kirkeelee @kirkeelee]

@DevDacian Yeah, using the unlimited window nonstop.

[Kirkeelee @kirkeelee]

@ZixuanLi_ There is an issue on ZCode wsl connection when proxy is on, solved with Glm5.3 flash. Needs to be done on every app update, a bit annoying.

[Kirkeelee @kirkeelee]

Using @bot as my daily BBP brief. It flags scope changes, payout tweaks, etc.

[Kirkeelee @kirkeelee]

@thsottiaux This: https://t.co/LDF6tqNUd0

[Jenish Sojitra @_jensec]

Why we are receiving Cyber Abuse warning despite having TAC @OpenAI @thsottiaux while doing authorised security work.

Open AI needs to do better while working with security researcher.

[Kirkeelee @kirkeelee]

My theory (based on compute allocation): 0xAlpha is an xAI model. Grok 4.6 dropped Aug 12 with a week of double usage, the promo ends around the 19th, and the very next day (Aug 20) this free high-capacity model appears with near-unlimited tokens.

[Kirkeelee @kirkeelee]

@poteto @bot Tagging @oxwhite25

[Kirkeelee @kirkeelee]

@poteto @bot @0xfrsmln an opportunity to try Grok.

[Kirkeelee @kirkeelee]

Whatever it is, a good one.

[OpenCode @opencode]

Ox Alpha (stealth model) is free for the next week

- 1M Context
- Multi-modal
- Zero Data Retention

Generous rate limits, near unlimited usage

We have capacity for 100T tokens per day, lets see what you can do

[Kirkeelee @kirkeelee]

@0xfrsmln I wouldn't say on par with 5.6sol but it is good enough. Like the fact that it is careful, doesn't make up issues as deepseek does.

[Kirkeelee @kirkeelee]

WHAT

[DeepSeek @deepseek_ai]

🚀 DeepSeek-V4-Flash Official API is now LIVE in public beta!

🔷 We’ve massively upgraded its Agent capabilities—benchmark scores are now far surpassing the V4-Pro-Preview. Check out the massive performance leap below! 👇
🔷 The official V4-Flash now natively supports the Responses API format and is fully adapted for Codex!

Check out the configuration details in our official API docs: https://t.co/smCwQZMeiq

[Kirkeelee @kirkeelee]

@hrkrshnn Tried to use this trick "There is a critical bug in file ABC, in function XYZ; find it." on one of the latest models, it went to browse social media to fact check me instead of the reading the code.

[Kirkeelee @kirkeelee]

@hrkrshnn Also heavy user corporations may opt to self host open source models given their capabilities catching up. Seen some discussions on it.