Michael_ (@MD_V_) profile picture
Germany
Joined June 2009
Photo by Michael_ (@MD_V_)Photo by Michael_ (@MD_V_)Photo by Michael_ (@MD_V_)
Anthropic filing for IPO with $80bn debt
113
510
73
8,008
251,031
Michael_ retweeted
So the fastest JSON serializer for .NET now is developed by CAPCOM as part of their game engine that powers Resident Evil / Biohazard I did not see that coming lol github.com/CAPCOM-TD-OSS/RED…
シリアライザーの話を聞きにきました
40
497
24
4,656
298,434
We’re disclosing HEIF Heist, a months-long investigation into libheif that allowed us to hack OpenAI, Slack, Meta, GitHub Ent, Rails, Next.js, ImageMagick, and many more. It was literally xkcd #234, one obscure image library beneath a huge number of apps. 🧵
48
427
66
3,233
561,991
Michael_ retweeted
Converts documents to audiobooks using Qwen3 TTS github.com/WhiskeyCoder/Qwen…
1
67
1
466
21,263
Michael_ retweeted
LiteLLM HAS BEEN COMPROMISED, DO NOT UPDATE. We just discovered that LiteLLM pypi release 1.82.8. It has been compromised, it contains litellm_init.pth with base64 encoded instructions to send all the credentials it can find to remote server + self-replicate. link below
296
2,191
605
9,182
5,922,649
A nice theater put on 😂
238
147
22
2,705
2,158,570
RIP fine-tuning ☠️ This new Stanford paper just killed it. It’s called 'Agentic Context Engineering (ACE)' and it proves you can make models smarter without touching a single weight. Instead of retraining, ACE evolves the context itself. The model writes, reflects, and edits its own prompt over and over until it becomes a self-improving system. Think of it like the model keeping a growing notebook of what works. Each failure becomes a strategy. Each success becomes a rule. The results are absurd: +10.6% better than GPT-4–powered agents on AppWorld. +8.6% on finance reasoning. 86.9% lower cost and latency. No labels. Just feedback. Everyone’s been obsessed with “short, clean” prompts. ACE flips that. It builds long, detailed evolving playbooks that never forget. And it works because LLMs don’t want simplicity, they want *context density. If this scales, the next generation of AI won’t be “fine-tuned.” It’ll be self-tuned. We’re entering the era of living prompts.
232
1,126
183
7,665
717,093
Michael_ retweeted
this github has a library of 90+ prompts for Nano Banana:
22
437
9
5,427
382,888
User named Krotter made a Paintball custom game mod using the new scripting system. They added a new weapon that shoots paint ball projectiles and a paint grenade. Even though the scripting is a bit limited, people's creativity is just getting started.
108
245
27
7,057
680,203
Two weeks ago, one of our chad engineers cooked so we released our IOMMU Restriction Enforcement, which marked the end of 2PC DMA attacks using IOMMU. This is where the device itself is contained in its own memory region and cannot read outside of it. No matter what you do, tampering with that would defeat the purpose of the 2PC "security" benefit. The biggest P2C devs, including devices like HPTT that cost $4500, have all either given up or are coping on theories of how they can get around it or try to resort to finding niche stupid things that gets them detected/banned lmao. It is a relatively simple area to cover, not a particularly problematic surface area. I have collected their tears for your enjoyment imgur.com/a/iommu-judgement-… and here is a video a 2PC DMA cheat dev has posted youtube.com/watch?v=IprU_G8M… This marks the end of 2PC DMA ATTACKS 2016-2025
108
65
47
1,153
182,535
Michael_ retweeted
shockingly good nuanced take 👏
121
96
14
2,168
159,555
New Anthropic Research: Project Vend. We had Claude run a small shop in our office lunchroom. Here’s how it went.
241
1,131
566
11,194
2,576,430
women who had c-section have a 3mm thick web connecting the scar to their shoulder. surgeons at mount sinai tracked 247 women post-surgery. 73% developed shoulder pain 2-7 years later. same side as the incision. no injury. just physics. 1/
every time you do a bicep curl 30% of your strength is leaking sideways and ends up in your leg. biomechanics researchers finally measured it only a couple of years back your muscles don't behave like you think. 1/
155
598
116
7,327
1,157,596
Michael_ retweeted
Introducing Dream Recorder — the magical bedside open-source device that plays your dreams back as cinematic reels. Build your own dreamrecorder.ai
308
368
359
5,302
1,056,645
🚀 The official Azure DevOps MCP is here! It includes an extensive set of tools you can use with GitHub Copilot to interact with Azure DevOps directly from Visual Studio Code. 🔗 Get to know more about the Azure DevOps MCP here: github.com/microsoft/azure-d… The tools range from working with Work Items and listing pull requests per project/repository to searching Wikis, and much more! #AzureDevOps #mcp #MVPBuzz
3
34
2
157
8,164
66
627
52
9,177
386,397
ASP #dotnet custom instructions file for GitHub Copilot Creating a copilot-instructions .md file allows us to give additional instructions to Copilot so it's responses can be specifically tailored for our repos. Below is an image excerpt of the current copilot instructions file for the dotnet/aspnetcore repo which was merged less than an hour ago. This feature is currently in preview and is currently supported for Copilot Chat in Visual Studio, VS Code and on the GitHub website. They are also supported for Copilot coding agent. What do you think?
8
24
1
183
14,434
Here's my insanely powerful o3 prompt — it guides o3 to think for WAY longer than it normally would, putting up to 5x more time and effort into solving your task: -- Ultra-deep thinking mode. Greater rigor, attention to detail, and multi-angle verification. Start by outlining the task and breaking down the problem into subtasks. For each subtask, explore multiple perspectives, even those that seem initially irrelevant or improbable. Purposefully attempt to disprove or challenge your own assumptions at every step. Triple-verify everything. Critically review each step, scrutinize your logic, assumptions, and conclusions, explicitly calling out uncertainties and alternative viewpoints. Independently verify your reasoning using alternative methodologies or tools, cross-checking every fact, inference, and conclusion against external data, calculation, or authoritative sources. Deliberately seek out and employ at least twice as many verification tools or methods as you typically would. Use mathematical validations, web searches, logic evaluation frameworks, and additional resources explicitly and liberally to cross-verify your claims. Even if you feel entirely confident in your solution, explicitly dedicate additional time and effort to systematically search for weaknesses, logical gaps, hidden assumptions, or oversights. Clearly document these potential pitfalls and how you've addressed them. Once you're fully convinced your analysis is robust and complete, deliberately pause and force yourself to reconsider the entire reasoning chain one final time from scratch. Explicitly detail this last reflective step. <task> {{PLACE_YOUR_TASK_HERE}} </task>
50
130
20
2,046
287,207