
I work at @EskomSePush ⚡
Cape Town, South Africa
Joined April 2009
- Tweets4.5K
- Following390
- Followers2.8K
- Likes6.5K
Dan 🦾 retweeted
Government’s new anti-misinformation chatbot… generates misinformation.
groundup.org.za/article/gove…
🇿🇦 Got questions about the South African Government? We’ve got answers!
Meet our new Government Chatbot — designed to help you find information and get informed. 🤖💬
👉 Click. Ask. Get informed!
🗣️ This is a pilot project, and your feedback matters! Try it out, share your experience and help us improve the service.
#GovZAUpdates #GovChatbot
🔗rsafacts.zoholandingpage.com…
卧槽,3090要可能要涨价了!!
因为老黄的刀法被人破解了
有人给3090解了个锁,同样限制在 360W,PyTorch 测试从 75 TFLOPS 跑到了 135 TFLOPS,提升约 80%!
你以为是超频,把功耗往上拉?
但他改的是显卡里控制运算速度的寄存器,这些寄存器平时有权限保护,正常改不了
他利用 GPU 启动固件的漏洞,赶在 NVIDIA 驱动加载之前拿到写入权限,把相关运算改成全速,再交给原版驱动使用
解开的主要是 FP16/BF16 输入、FP32 累加这类 AI 运算的速度限制 但测试算力多了 80%!!
4090 也有分支项目做到了:同类运算的指令吞吐,从约 165 TFLOPS 提升到 331 TFLOPS,接近翻倍
项目地址,附上👇
github.com/Zagitta/gpu-unloc…
Dan 🦾 retweeted
Capetonians!
There will be some loud noise in the sky above Cape Town and the West Coast for the next week, as two 2 Sqn Gripen jets of the SAAF take part in Exercise IBSAMAR IX, between the navies of South Africa, Brazil and India.
Please watch your pets.
📷 Sebastian Horrell
You don't have to commit your entire life to optimism, just try it on for a week. Start thinking "what if it all worked out?" and "what a time to be alive!". Then compare your state of mind to the week before filled with pessimism and doomer nonsense.
Dan 🦾 retweeted
Load shedding incoming for Soccer City… @EskomSePush
Dan 🦾 retweeted
Coded a funny thing today while we were having some burgers and beers, see how low your flyover can go
flappylink.com
Yesterday me and my friends talked about the Dead Internet Theory
If nobody asks questions anymore on Stack Overflow, and sites like Reddit are now taken over by AI reply bots to promote brands, as well as AI reply bots on here, there is no real content anymore on the internet
And then there is no fresh training data anymore
I thought about something like, how would you find out the best outdoor action camera? Before I would search :
site:reddit.com best outdoor action camera
But now I just ask my AI which is the best
The problem is if nobody creates new content anymore and just asks their AI every question, the AI has nothing to train on anymore
And then I thought so what will happen? Will you get companies with giant warehouses that just buy stuff and review it with manually or with humanoid robots? Will you get humanoid robots backpacking in South East Asia to get life experience as training data? Or will we all train it by giving it access to our AI chats and user data?
I don't know but I think it's likely the web will dead in the future and AI companies will have to find training data elsewhere
Dan 🦾 retweeted
When loadshedding ended, a lot of people uninstalled ESP.
We assumed loadshedding is over, everyone stops it and the app dies.
It didn’t.
Have you uninstalled ESP?
Dan 🦾 retweeted
I co-founded pfSense. For the last year I've been building its successor.
If you run pfSense today, you already know the reasons to look around: development you can't influence, a CE edition that feels like an afterthought, FreeBSD driver roulette on modern hardware, and a config workflow where one bad apply on a remote box means a drive.
nfSensei is my answer. Built from scratch in Rust, on Linux, and designed around the things pfSense users actually complain about.
Your config.xml imports. There's an importer that reads your pfSense config, shows you exactly what maps over and what needs attention, then applies it. You don't start from zero.
You can't brick it from the couch. Changes stage as a candidate, diff before apply, validate through the real engines before anything is written, and auto-roll-back if you don't confirm in time. If a config ever fails at boot, the box falls back to the last good one on its own.
The hardware works. Linux base means modern NICs and drivers just work — and the fast path compiles your rules to XDP at 40Gbps.
Automation is native, not scraped. Everything the UI does is a documented API call — about 1,140 of them, with a built-in explorer. Your Ansible finally gets a real interface.
The VPNs are current. WireGuard, IPsec, Tailscale, and self-hosted mesh — your own control plane, your keys — plus post-quantum key exchange where it counts.
The experimental stuff has its own wing. Thirty-plus Labs features behind toggles: WAN bonding that fuses multiple cheap uplinks through a $5 VPS into one resilient pipe, per-flow SLA telemetry with tamper-evident audit chains, GeoDNS that steers traffic by live RTT and load, application-aware QoS, config push to a whole fleet of remote nodes, and an AI assistant on the box that reads your actual interfaces and logs using local models. Toggles are per-browser and can't touch your running config — flip things on, break them, tell me about it. Oh, and there is much more to mention here!
Self-hosted, on your hardware, no cloud account, no subscription.
It's NOW IN ACTIVE BETA (previously alpha) with about 40 testers, and bug reports typically get fixed in days. I want more people who know what pfSense does well and can tell me exactly where nfSensei falls short.
If you are interested in testing please email me: [email protected]. Tell me about your pfSense setup and I'll get you access.
Note: Affiliates and employees of Netgate are not invited.
Dan 🦾 retweeted
vLLM v0.22.0 is out! 459 commits from 230 contributors (63 new). 🎉
Highlights: DeepSeek V4 hardening (NVFP4 fused MoE, full+piecewise CUDA graph, ROCm support), experimental Rust frontend in-Tree, batch-invariant Cutlass FP8 (28.9% lower e2e latency), Model Runner V2 advances, multi-tier KV cache offloading.
Thread 👇
🤖 Made with AI
Dan 🦾 retweeted
llama.cpp now has an official website: llama.app
Our goal is to make local AI accessible to everyone, and improving the user experience is a big part of that. On the new landing page you’ll find a single-line cross-platform installer. The installation provides a single unified `llama` entrypoint which you can use to run/serve models and interface with 3rd-party agentic applications.
While oriented towards simplified user experience, the new `llama` application also provides all the advanced functionality of the existing llama.cpp tooling with which experienced users are already familiar. Also note that all GGUF models that you might have already downloaded with llama.cpp in the past will be automatically available to use without downloading again (they are stored in the common HF cache on your machine).
We have many improvements in the pipeline both at the UX and at the engine level and we plan to iteratively ship new things over the coming months. One of the main focuses will be seamless integration with local-friendly 3rd-party agents (such as Pi). In the meantime, we’ll continue to listen for feedback from the community and adjust accordingly, so keep letting us know what you think and need.


























