AI Unfiltered

9 Min Read

onApril 28, 2026

DeepSeek AI Releases Manifold-Constrained Hyper-Connections Architecture—New Transformer Design Cuts Overfitting and Computational Cost Through Dynamic Neuron Links Within Mathematical Manifolds

A Beijing lab just made every transformer architecture from 2017-2025 look like a brute-force approximation. Two competing labs shipped…

Discover More

Natural Language Processing

8 Min Read

Artur Markus

onApril 27, 2026

No Breaking News Found in Natural Language Processing Category (Past 7 Days)

The quietest week in NLP during 2026 tells us more about where the field is headed than any single announcement would. When the news cycle…

Discover More

9 Min Read

Artur Markus

onApril 26, 2026

Zapier Agents Hit General Availability with Multi-Step Memory—All Five Major Workflow Platforms Ship Agent Builders in Single Quarter

Five competitors shipping the same feature within 90 days isn’t a product launch—it’s an industry admitting that rule-based…

Discover More

10 Min Read

Artur Markus

onApril 25, 2026

Allen Institute’s MolmoWeb 8B Beats GPT-4o on Web Navigation—First Open-Weight Agent to Outperform Proprietary Models Across 4 Major Benchmarks

An 8-billion parameter open-source model just outperformed GPT-4o at navigating the web. Allen Institute released everything—training code,…

Discover More

9 Min Read

Artur Markus

onApril 24, 2026

Alibaba’s Qwen Hits 700 Million Downloads on Hugging Face—Overtakes Meta’s Llama as World’s Most Popular Open-Source AI Model

A Chinese AI model just quietly dethroned Meta’s Llama as the world’s most downloaded open-source system—and December’s…

Discover More

Productivity with AI

7 Min Read

Artur Markus

onApril 23, 2026

No Breaking News Found in Productivity with AI Category (Past 7 Days)

I searched 200+ sources for breaking AI productivity news this week. Found exactly zero announcements worth your attention—and that silence…

Discover More

10 Min Read

Artur Markus

onApril 23, 2026

Palantir’s Maven Smart System Running on Anthropic’s Claude Powers 11,000+ US Strikes in Iran—DoD Designates It Official Programme of Record with 25,000+ Military Accounts Deployed

The Pentagon just made AI-powered killing permanent infrastructure. Palantir’s Maven Smart System, running on Anthropic’s Claude,…

Discover More

9 Min Read

Artur Markus

onApril 21, 2026

Cursor Raises $2 Billion at $50 Billion Valuation—Fastest B2B Software Company to Hit $1 Billion ARR, Outpacing Slack, Zoom, and Snowflake

A three-year-old coding editor just became the fastest B2B software company to reach $1 billion ARR—beating Slack’s record by roughly…

Discover More

9 Min Read

Artur Markus

onApril 19, 2026

Apiiro CLI Launches April 9 Turning AI Coding Assistants Into Security Engineers by Rewriting Developer Prompts in Real Time

Your AI coding assistant now has a shadow editor—one that rewrites your prompts before the AI model sees them, injecting security constraints…

Discover More

9 Min Read

Artur Markus

onApril 18, 2026

35 State Attorneys General Launch Coordinated Investigation into xAI’s Grok for Generating Nonconsensual Intimate Images and CSAM

The largest coordinated state enforcement action against an AI company just launched—and it targets content safety failures that most…

Discover More

8 Min Read

Artur Markus

onApril 17, 2026

Anthropic’s Constitutional Classifiers++ Cut Jailbreak Success Rate from 86% to 4.4%—Only 1 Universal Jailbreak Found in Bug Bounty Testing

Anthropic just compressed what should have been years of AI safety progress into one architecture update—blocking 95% of jailbreaks for 1%…

Discover More

9 Min Read

Artur Markus

onApril 16, 2026

Anthropic Delays Mythos AI Model After It Autonomously Exploited Tens of Thousands of Software Vulnerabilities with 80% Success Rate

Anthropic just became the first major AI lab to delay a frontier model because it got too good at breaking into systems. Mythos achieved an…

Discover More

9 Min Read

Artur Markus

onApril 15, 2026

Meta Llama 4 Launches April 5 with 109B-Parameter Scout Model—17B Active Parameters Fit on Single H100, Outperform GPT-4.5 and Claude 3.7 on STEM Benchmarks

Meta just made frontier AI fit on one GPU—and simultaneously banned 450 million Europeans from using it.

Discover More

DeepSeek AI Releases Manifold-Constrained Hyper-Connections Architecture—New Transformer Design Cuts Overfitting and Computational Cost Through Dynamic Neuron Links Within Mathematical Manifolds

No Breaking News Found in Natural Language Processing Category (Past 7 Days)

Zapier Agents Hit General Availability with Multi-Step Memory—All Five Major Workflow Platforms Ship Agent Builders in Single Quarter

Explore Trending Topics

Subscribe to my Blog