Gurveer (@gurvrx) header
Gurveer (@gurvrx) profile picture

I'm building @aurorallm — the cheapest way to call every LLM. Open-source. Auto-routes to the cheapest capable model. Cut your bill.

earth
Joined January 2025
Photo by Gurveer (@gurvrx)Photo by Gurveer (@gurvrx)Photo by Gurveer (@gurvrx)Photo by Gurveer (@gurvrx)Photo by Gurveer (@gurvrx)Photo by Gurveer (@gurvrx)
CVE-2026-83844: Remote Code Execution in Microsoft Paint. cve.org/CVERecord?id=CVE-202…
1
2
232
Mspaint affected from v11.0.0.0 to before 11.2605.91.0
1
118
🩸 Ivanti Connect Secure = ROOT in the wild. 3 chained zero-days, actively exploited by UTA0576: CVE-2026-104935 (unauth RCE 8.6) CVE-2026-111357 (priv-esc to root 8.8) ← CISA KEV, patch by 11/2 CVE-2026-118527 (file upload, deadline 11/13)
1
1
185
कांग्रेस के समय में IDBI बना - देश का उद्योग खड़ा करने के लिए। मोदी के समय में पहले बोली देखी गई, फिर दाम तय हुआ। कांग्रेस के time पे बैंक बनते थे। मोदी के time में बिकते हैं। मैं IDBI बैंक के कर्मचारियों के साथ खड़ा हूँ।
1,419
11,524
440
35,514
1,560,162
MYTHOS production tool verified. Live Chrome CDP attach works. Session persists, no re-auth needed. 🗡️
8
441
Testing the MYTHOS X-poster. Live automation check. #MYTHOS
2
303
not to be dramatic but one good conversation could change the trajectory of my entire day
10
415
DeepSeek v4.1 flash High vs mimimax m3.1 flash High
1
399
Critical Vulnerabilities PHP CVE-2026-109618 / CVE-2026-109619, fixed in 8.4.27. Four services, one command: - 8.4.26 target - 8.4.27 control - chain A: session injection + LFI - chain C: handler mismatch 15 Suricata, 5 Sigma, 7 ModSecurity, OpenResty. Replay-tested on pcaps.
7
3
528
Whole lab is docker compose — vulnerable target, patched control that must fail, two patch-independent chains. Detection fixtures generate their own pcaps, so you can validate the rules without a target.
21
Why 1001 form parts before the interesting ones? To blow past max_input_vars and force the parser into the OOM path. The prime is a 12MB multipart body with a NUL-byte PHPSESSID smuggled in as a form field — not a cookie. PHP truncates the key at the NUL.
21
Second request leaks libphp through a 56-byte HashTable page. Third request: a 167-byte live zend_string laid out as a fake HashTable, with pDestructor pointing at zif_system. Command budget after the struct: 14 bytes. `id>/out/p1` fits. Barely.
26
The bug: session.upload_progress + memory exhaustion during request startup. PHP hands back uninitialized heap in the session cookie. A 26-char session id comes back as ~4KB of live pointers. First leak = where the heap is. That's ASLR dead.
40
I wrote detection rules for a PHP RCE, then measured whether they fire. Suricata 8.0.7 inspects ~4KB of a request body. The exploit puts its payload at ~90KB. Every deep-content rule I wrote was decorative. 🧵
293
One rule I gave the model for this one: Nothing in the scene is drawn or animated in code. No canvas snow. No SVG weather. No fake light. Just a photo plate, looping video on identical cover rules, and a measured 1546×1017 canvas that scales to any screen. prompt in comments
🤖 Made with AI
1
3
254
I gave an AI 1,400 words of instructions and almost no freedom. Exact easing curve. Exact breakpoints. Exact pixel where the headline is allowed to sit. It built a 3D photo sphere with an intro film and a custom cursor. The prompt is free. Repo in the reply.
1
3
164