←
SeanG945 posts
SeanG avatar

SeanG

@AnotherWayIn
◷ Joined December 2009▧ 85 Photos and videos
clem 🤗 avatar
⟳ SeanG retweeted
clem 🤗@ClementDelangue·Nov 17

Introducing Qwen3.8-27B-pi-adapted 🤗 We adapted Qwen's model specifically for the pi coding agent using reinforcement learning and a LoRA adapter. The results, measured as pass@1: 🚀 SWE-bench Pro (Hard 51): 29.4% → 62.7% 💻 Terminal-Bench 2.1: 39.5% → 55.8% ⚡ 55–62% fewer tool calls per task This is the power of open models: you can adapt them to YOUR tools and YOUR use cases. Let’s build! 💛 Open weights: huggingface.co/harness-hub/Q… Methodology: huggingface.co/blog/harness-…

Post attachment
SeanG avatar
SeanG@AnotherWayIn·20h

it'll be easier to just bookmark updetector next year 😅

SeanG avatar
SeanG@AnotherWayIn·Nov 16

My early tests show Qwen 4 has the exact same cyber capabilities as DS 1017. Have we now hit the ceiling for China export controls?

SeanG avatar
SeanG@AnotherWayIn·Nov 15

Very soon, i suspect we’ll realise we don’t need all that compute after all.

SeanG avatar
SeanG@AnotherWayIn·Nov 15

Having my Dot work on stuff with me through DSH while also be sending other cheap tasks off to my sparks directly is a real level up

SeanG avatar
SeanG@AnotherWayIn·Nov 14

Nearly 3hrs into my first @foofighters concert and I’m blown away at how good they are. Can’t believe I’ve missed the last 30yrs of them. Hands down the best gig I’ve been to. Dave has soo much energy 🤘

SeanG avatar
SeanG@AnotherWayIn·Nov 14

The new reality is here. AI brings lots more attacks, which also means more patching - which is often more dangerous

SeanG avatar
SeanG@AnotherWayIn·Nov 13

FYI if you goto codex plugins and search for daybreak (if you have access) you can use the new vuln scan plugin with $500 credits included

Post attachment
SeanG avatar
SeanG@AnotherWayIn·Nov 13

just removing the word 'hmm' from DeepSeek's training could save so many trees

Post attachment
International Cyber Digest avatar
⟳ SeanG retweeted
International Cyber Digest@IntCyberDigest·Nov 12

‼️ A hacker says an 8GB Nvidia crypto-mining card hides 80GB of memory. She unlocked it, and in her tests it outran Nvidia's A100 AI chip. The CMP 170HX uses the same chip as the A100, with the extra memory locked off by fuses. Kate broke the lock by glitching its voltage with a homemade STM32 board that she says an AI model designed. That switched on up to 126 compute units, more than the 108 on any A100. It hit 85.9 TFLOP/s in FP16 against 72.5 for a retail A100, with all 80GB usable. Each card took 1 to 48 hours of glitching. One of her 10 only gave up 64GB.

Post attachmentPost attachment
Pix🔎 avatar
⟳ SeanG retweeted
Pix🔎@PixBuilds·Nov 9

How To Tell If Your Girlfriend Is A Spy:

Post attachment