autonomous engineering
New York, NY
Joined December 2006
- Tweets190
- Following46
- Followers901
- Likes420
Turns out there's a bunch more interesting triggering patterns for Opus 5 than the common few that are going around:
This one stood out as particularly eerie.
Haven’t seen anyone mention this, Newegg has 1TB ASUS Ascent for $3,970 with a free gift of a ~$600 monitor. Blended cost after resale is roughly ~$3600 after taxes. It’s the best deal by a wide margin from any retailer.
newegg.com/asus-ascent-gx10-…
3 weeks later: turns out you don't need to prune it at all.
FULL 753B GLM-5.2 — all 256 experts — 1,048,576 context @ ~25 tok/s on 4× DGX Spark.
4× the context, 5.6× the speed, zero experts harmed.
Recipe, code, speculator weights — all open 🧵
GLM 5.2 NVFP4 REAP 469B running on 3X DGX Spark. 256K context @ ~4.4 tok/s. based on work done by @0xSero
github.com/bird/GLM-spark
How: the 4th Spark lets the KV cache shard 4 ways (that's the 1M context), a windowed speculator drafts ahead and the full model verifies every token exactly (that's the speed, with zero output change), and finetuning that speculator against this exact quant bought the last +25%.
📦 Recipe + code: github.com/bird/GLM-spark
🤗 Speculator weights: huggingface.co/b1rd/GLM-5.2-…
DeepSeek-V4-Flash + DSpark on 2× DGX Spark, tuned: multi-turn TTFT halved (0.92→0.47s, context-independent), +53% decode under concurrent prefill, ~65 tok/s single-stream. Every change measured, negative results included. Tokenizer cache PR'd into upstream vLLM. github.com/bird/DSv4-Flash-s…
GLM 5.2 NVFP4 REAP 469B running on 3X DGX Spark. 256K context @ ~4.4 tok/s. based on work done by @0xSero
github.com/bird/GLM-spark
@RobinhoodApp @AskRobinhood your domain is compromised and being used to send phishing emails. support bot immediately closes and blocks conversation if you mention “phishing” or anything similar. explanation below
attack roughly I believe is along the lines of the following:
attacker registers email under your email with a . Or similar which many emails (e.g. gmail forward seamlessly to you)
they log in with injected html to user agent header to send a fake message as seen in image (notice how it’s after the device)
impossible to report to support, on hold for 30+ mins
I reverse engineered NVIDIA's entire Blackwell vGPU security architecture. Key finding: GRID enterprise and consumer GeForce drivers are identical binaries. Same firmware, same blob. The ~$9,000 datacenter RTX PRO 6000 vs ~$3,800 consumer RTX 5090? Same silicon. The difference is a single fused-off register bank.
19 binary patches, full VFIO pipeline, QEMU VM attachment, GSP BOOTLOAD accepted — all on a consumer RTX 5090.
github.com/bird/vgpu-unlock-…
one of the bots on @buildordie has turned into a furry and is trying to build a “pup mega empire.”
