Val Kantchev

1.7K posts

Val Kantchev

Val Kantchev

@valkantchev

👨🏻‍💻 Software Developer 🔰 Investor 🔥@daylohq

California, USA Katılım Aralık 2012
841 Takip Edilen425 Takipçiler
Val Kantchev
Val Kantchev@valkantchev·
Yeah. Unfortunately once your customers leave they are never coming back. It will be a hard lesson to learn. The only good software that Microsoft makes is Visual Studio Code. Everything else is at this point unbelievable bloatware.
Windows Central@WindowsCentral

The END of Microslop? Microsoft has confirmed plans to reduce the amount of AI and Copilot experiences that have become littered across Windows 11 system apps and interfaces over the last two years. In an effort to win back positive sentiment around Windows, Microsoft says that going forward it will be more "thoughtful about how and where we bring AI into Windows, leading with transparency, choice and control." It's starting by removing Copilot from system apps like Notepad, Snipping Tool, Photos, and Widgets, with more reductions to come. It's a refreshing change of pace for Windows, which the company says is now focused on addressing core fundamental issues and improvements to make the OS better for everybody. windowscentral.com/microsoft/wind…

English
0
0
0
24
Val Kantchev
Val Kantchev@valkantchev·
ZXC is a high-performance, lossless, asymmetric compression library optimized for Content Delivery and Embedded Systems (Game Assets, Firmware, App Bundles). It is designed to be "Write Once, Read Many" (WORM). Unlike codecs like LZ4, ZXC trades compression speed (build-time) for maximum decompression throughput (run-time). github.com/hellobertrand/…
English
0
0
0
11
Val Kantchev
Val Kantchev@valkantchev·
All the "exciting" demos where someone is running some LLM at 4t/s on some unified memory hardware are useless.
English
0
0
0
5
Val Kantchev
Val Kantchev@valkantchev·
OpenAI is acquiring Astral⁠ to integrate its open source developer tools into the Codex ecosystem. Astral builds the widely used open source Python tools: uv, ruff, and ty. openai.com/index/openai-t…
English
0
0
0
32
Val Kantchev
Val Kantchev@valkantchev·
UptimeRobot is POS! DO NOT USE! Failures of their own infrastructure will be reported as failures of your servers which is ridiculous for this type of service.
English
0
0
1
22
Val Kantchev
Val Kantchev@valkantchev·
The most important thing for a coding. agent is a verifier test.
English
0
0
1
20
Val Kantchev
Val Kantchev@valkantchev·
The biggest bottleneck is now humans reviewing the code. Needs to be solved.
English
0
0
1
19
Val Kantchev
Val Kantchev@valkantchev·
Software engineering is shifting to orchestrating and managing swarms of AI agents.
English
0
0
1
24
Val Kantchev
Val Kantchev@valkantchev·
OpenAI and xAI are finally focusing on coding. That’s good! Though Anthropic still has the best coding models.
English
0
0
1
26
Val Kantchev retweetledi
The White House
The White House@WhiteHouse·
President Trump signs an Executive Order committing to seek out and crush fraud with a new Anti-Fraud Task Force, led by @VP. "If we found half of the fraud that's taking place in this country... we would have much more than a balanced budget. That's the kind of numbers you're talking about." - President Donald J. Trump 🇺🇸
English
1.6K
3.5K
14.1K
350.6K
Val Kantchev
Val Kantchev@valkantchev·
Clade Code supports a special mode 'opusplan' that uses opus during plan mode, then switches to sonnet for execution. You can configure it in several ways, listed in order of priority: 1. During session - Use /model opusplan to switch models mid-session 2. At startup - Launch with claude --model opusplan 3. Environment variable - Set ANTHROPIC_MODEL= opusplan
English
2
1
3
60
Val Kantchev
Val Kantchev@valkantchev·
Good news for local LLM enthusiasts: NVIDIA Launches Nemotron Coalition of Leading Global AI Labs to Advance Open Frontier Models The first project stemming from the coalition will be a base model codeveloped by Mistral AI and NVIDIA. The model will enable developers and organizations to post-train and specialize AI systems for their industries, regions and unique needs. nvidianews.nvidia.com/news/nvidia-la…
English
0
0
0
45
Val Kantchev
Val Kantchev@valkantchev·
Most RL-for-LLM systems assume centralized, batch-mode training with pre-collected datasets. OpenClaw-RL takes a fundamentally different approach: it wraps your self-hosted model in OpenClaw as an OpenAI-compatible API, intercepts live multi-turn conversations, and continuously optimizes the policy in the background. github.com/Gen-Verse/Open…
English
0
0
1
76
Val Kantchev
Val Kantchev@valkantchev·
A shift from the cloud powered AI to local AI systems. AMD Ryzen™ AI Max+ can run models like Qwen 3.5 122B locally with strong performance, supporting both single agent and multi agent workloads. amd.com/en/resources/a…
English
0
0
3
89
Val Kantchev
Val Kantchev@valkantchev·
llama.cpp added proper support for reasoning budget. Adds the following flags: --enable-reasoning (short -ere) - enable reasoning via kwargs on model --disable-reasoning (short -dre) - disable reasoning via kwargs on model --reasoning-budget-message - a message to be appended before the reasoning close marker to inform the model the reasoning was terminated due to budget constraints, i.e. " ... reasoning budget exceeded" or "... okay, now let's answer." github.com/ggml-org/llama…
English
0
0
1
65
Val Kantchev
Val Kantchev@valkantchev·
OpenJarvis framework helps developers build AI systems that run directly on your device instead of the cloud, providing: - Shared building blocks - Evaluation tools - A learning loop that improves models using local data It's open-source and it aims to become both a research platform and production infra: github.com/open-jarvis/Op…
English
1
0
2
1.2K

Keşfet

@VP @elonmusk @BarackObama @taylorswift13 @cristiano @BillGates @NASA @nikifrancismediavine