📋 Log for 2026-03-26

😄 Joke of the Day

How does a French skeleton say hello? Bone-jour.

Category: dad

YouTube Summaries

[The AI Job Market Split in Two. One Side Pays $400K and Can't Hire Fast Enough.](https://www.youtube.com/watch?v=4cuT-LKcmWs)

Channel: NateBJones

Summary:

• Here's a summary of the video "The AI Job Market Split in Two. One Side Pays 00K and Can't Hire Fast Enough.":

Key Takeaways

• The AI job market is characterized by a "K-shaped" distribution, meaning there's a significant divide between highly in-demand professionals and others, rather than a general scarcity of AI jobs.
• Despite a high number of open AI roles (stated as 3.2 jobs per qualified candidate), employers face immense difficulty filling positions due to a critical skills gap among applicants.
• Acquiring a specific set of seven learnable skills can dramatically increase an individual's value in the AI job market, leading to significant career opportunities.

Main Arguments

• The prevailing narrative of scarce AI jobs is misleading; the reality is an abundance of roles coupled with a severe mismatch between employer needs and candidate qualifications.
• Employers are actively seeking candidates who possess practical, production-ready skills that enable them to effectively build, deploy, and manage AI systems, not just those with theoretical knowledge or basic AI credentials.
• The seven highlighted skills are crucial for bridging the gap between what companies require and what candidates currently offer, making them highly valuable for career advancement.

Notable Points (Paraphrased from the transcript/description)

• "Your AI credentials don't matter; your skills do." (Implied by the substack title and focus on practical skills).
• "The gap between what employers need and what candidates offer has never been wider or more correctable."
• "3.2 AI jobs exist for every qualified candidate."
• Employers are interviewing hundreds of candidates yet remain unable to fill critical roles.

Important Nuances

• K-Shaped Market: This describes a market where some segments (highly skilled AI professionals) are thriving, experiencing high demand and compensation, while others may struggle, leading to an uneven economic landscape.
• Seven Learnable Skills: The video emphasizes specific, actionable skills driving demand:
• Specification Precision: The ability to clearly and precisely instruct AI models.
• Evaluation and Quality Judgment: The critical skill of assessing AI outputs for accuracy, relevance, and overall quality.
• Multi-Agent Decomposition and Delegation: The capacity to break down complex tasks for AI agents and manage their interactions effectively. This also includes matching project scope to the capabilities of the "agentic harness" (the system an agent operates within).
• Failure Pattern Recognition: The ability to diagnose and understand why AI systems fail, crucial for production environments. The video mentions six specific failure types to diagnose.
• Trust and Security Design: Building AI systems that are secure, reliable, and can be trusted.
• Context Architecture: Designing how AI models access, manage, and utilize contextual information.
• Cost and Token Economics: Understanding and managing the financial implications and resource usage (tokens) of AI models.
• Production-Ready Builders: The focus is on candidates who can deliver functional AI solutions in real-world applications, not just those with academic or theoretical understanding.

Published: 2026-03-26T15:59:47+00:00

Latest OpenRouter Models

Google: Gemma 4 26B A4B (google/gemma-4-26b-a4b-it)

Gemma 4 26B A4B IT is an instruction-tuned Mixture-of-Experts (MoE) model from Google DeepMind. Despite 25.2B total parameters, only 3.8B activate per token during inference — delivering near-31B quality at a fraction of the compute cost. Supports multimodal input including text, images, and video (up to 60s at 1fps). Features a 256K token context window, native function calling, configurable thinking/reasoning mode, and structured output support. Released under Apache 2.0.

Published: 03/04/2026

https://openrouter.ai/google/gemma-4-26b-a4b-it

Google: Gemma 4 31B (google/gemma-4-31b-it)

Gemma 4 31B Instruct is Google DeepMind's 30.7B dense multimodal model supporting text and image input with text output. Features a 256K token context window, configurable thinking/reasoning mode, native function calling, and multilingual support across 140+ languages. Strong on coding, reasoning, and document understanding tasks. Apache 2.0 license.

Published: 02/04/2026

https://openrouter.ai/google/gemma-4-31b-it

Qwen: Qwen3.6 Plus (free) (qwen/qwen3.6-plus)

Qwen 3.6 Plus builds on a hybrid architecture that combines efficient linear attention with sparse mixture-of-experts routing, enabling strong scalability and high-performance inference. Compared to the 3.5 series, it delivers major gains in agentic coding, front-end development, and overall reasoning, with a significantly improved “vibe coding” experience. The model excels at complex tasks such as 3D scenes, games, and repository-level problem solving, achieving a 78.8 score on SWE-bench Verified. It represents a substantial leap in both pure-text and multimodal capabilities, performing at the level of leading state-of-the-art models.

Published: 02/04/2026

https://openrouter.ai/qwen/qwen3.6-plus

Free Models Catalog

Arcee AI: Trinity Large Preview (free) (arcee-ai/trinity-large-preview)

Trinity-Large-Preview is a frontier-scale open-weight language model from Arcee, built as a 400B-parameter sparse Mixture-of-Experts with 13B active parameters per token using 4-of-256 expert routing.

It excels in creative writing, storytelling, role-play, chat scenarios, and real-time voice assistance, better than your average reasoning model usually can. But we’re also introducing some of our newer agentic performance. It was trained to navigate well in agent harnesses like OpenCode, Cline, and Kilo Code, and to handle complex toolchains and long, constraint-filled prompts.

The architecture natively supports very long context windows up to 512k tokens, with the Preview API currently served at 128k context using 8-bit quantization for practical deployment. Trinity-Large-Preview reflects Arcee’s efficiency-first design philosophy, offering a production-oriented frontier model with open weights and permissive licensing suitable for real-world applications and experimentation.

Published: Tue, 27 Jan 2026 22:24:30 GMT

LiquidAI: LFM2.5-1.2B-Instruct (free) (liquid/lfm-2.5-1.2b-instruct)

LFM2.5-1.2B-Instruct is a compact, high-performance instruction-tuned model built for fast on-device AI. It delivers strong chat quality in a 1.2B parameter footprint, with efficient edge inference and broad runtime support.

Published: Tue, 20 Jan 2026 16:45:21 GMT

Venice: Uncensored (free) (cognitivecomputations/dolphin-mistral-24b-venice-edition)

Venice Uncensored Dolphin Mistral 24B Venice Edition is a fine-tuned variant of Mistral-Small-24B-Instruct-2501, developed by dphn.ai in collaboration with Venice.ai. This model is designed as an “uncensored” instruct-tuned LLM, preserving user control over alignment, system prompts, and behavior. Intended for advanced and unrestricted use cases, Venice Uncensored emphasizes steerability and transparent behavior, removing default safety and alignment layers typically found in mainstream assistant models.

Published: Wed, 09 Jul 2025 21:02:46 GMT

Google: Gemma 3n 2B (free) (google/gemma-3n-e2b-it)

Gemma 3n E2B IT is a multimodal, instruction-tuned model developed by Google DeepMind, designed to operate efficiently at an effective parameter size of 2B while leveraging a 6B architecture. Based on the MatFormer architecture, it supports nested submodels and modular composition via the Mix-and-Match framework. Gemma 3n models are optimized for low-resource deployment, offering 32K context length and strong multilingual and reasoning performance across common benchmarks. This variant is trained on a diverse corpus including code, math, web, and multimodal data.

Published: Wed, 09 Jul 2025 15:28:24 GMT

Qwen: Qwen3 4B (free) (qwen/qwen3-4b)

Qwen3-4B is a 4 billion parameter dense language model from the Qwen3 series, designed to support both general-purpose and reasoning-intensive tasks. It introduces a dual-mode architecture—thinking and non-thinking—allowing dynamic switching between high-precision logical reasoning and efficient dialogue generation. This makes it well-suited for multi-turn chat, instruction following, and complex agent workflows.

Published: Wed, 30 Apr 2025 16:38:24 GMT

Robot Technology

🤖 30 Years Ago, Robots Learned to Walk Without Falling

When you hear the term humanoid robot , you may think of C-3PO , the human-cyborg-relations android from Star Wars . C-3PO was designed to assist humans in communicating with robots and alien species. The droid, which first appeared on screen in 1977, joined the characters on their adventures, walking, talking, and interacting with the environment like a human. It was ahead of its time. Before the release of Star Wars , a few androids did exist and could move and interact with their environment,...

Source: spectrum.ieee.org • Published: Wed, 25 Mar 2026 18:00:05 +0000

Read more

🤖 Sweet success: How a world-renowned candy manufacturer in Germany scaled palletizing across multiple lines

Automation rarely starts with a full factory transformation. More often, it begins with a single line. One challenge, one opportunity, one team ready to try something different. What matters is what happens next.

Source: blog.robotiq.com • Published: Thu, 26 Mar 2026 13:48:52 GMT

Read more

🤖 Deepfake X-rays are so real even doctors can’t tell the difference

Deepfake X-rays created by AI are now convincing enough to fool both doctors and AI models. In tests, radiologists had limited success identifying fake images, especially when they didn’t know they were being shown. This opens the door to risks like fraudulent medical claims and tampered diagnoses. Experts say stronger safeguards and detection tools are critical as the technology advances.

Source: www.sciencedaily.com • Published: Thu, 26 Mar 2026 06:42:12 EDT

Read more

Good News

First Quantum Battery Prototype Marks Big Step for Technology Expected to Change the World

Australian researchers have developed and tested the world’s first quantum battery. Their prototype is far from anything that will be a perspective power source in an EV or storage facility, but the experiment revealed some important directions for future research. A theoretical concept since 2013, the prototype was charged wirelessly with a laser, one of […] The post First Quantum Battery Prototype Marks Big Step for Technology Expected to Change the World appeared first on Good New...

Published: Wed, 25 Mar 2026 18:30:23 +0000

Read more

Marine National Park Bigger Than Texas Created by Departing Chilean President Around Pacific Islands

Just days before he left office, former Chilean President Gabriel Boric signed into law protections for the remote Juan Fernández Archipelago and a gargantuan swath of the surrounding sea. The protections connect the archipelago with the existing Nazca-Desventuradas marine parks, and total 386,000 square miles, an area the size of Venezuela amounting to 50% of […] The post Marine National Park Bigger Than Texas Created by Departing Chilean President Around Pacific Islands appeared first on...

Published: Thu, 26 Mar 2026 11:00:02 +0000

Read more

Good News in History, March 26

51 years ago today, the Biological Weapons Convention, (BWC) entered into force. It’s considered to have established a strong global norm against biological weapons reflected in the treaty’s preamble, which states that the use of biological weapons would be “repugnant to the conscience of mankind.” It is also demonstrated by the fact that not a […] The post Good News in History, March 26 appeared first on Good News Network .

Published: Thu, 26 Mar 2026 07:00:00 +0000

Read more

Brazil’s new law blames platform design for harming kids, not parents

BY THE OPTIMIST DAILY EDITORIAL TEAM If you have ever lost an hour to a video feed you never meant to open, you understand what Brazil just decided to make illegal for children. The Digital Statute of Children and Adolescents took effect in Brazil last week, and what makes it different from most online safety […] The post Brazil’s new law blames platform design for harming kids, not parents first appeared on The Optimist Daily: Making Solutions the News .

Published: Thu, 26 Mar 2026 00:00:17 +0000

Read more

Giant sequoia clones from 3,000-year-old trees are taking root in Detroit

BY THE OPTIMIST DAILY EDITORIAL TEAM In California’s Sierra Nevada, giant sequoias have stood for millennia. The largest trees top 300 feet, live past 3,000 years, and are among the biggest living things on Earth by mass. Now, clones of specific ancient trees are being planted in Detroit. Volunteers with Archangel Ancient Tree Archive and […] The post Giant sequoia clones from 3,000-year-old trees are taking root in Detroit first appeared on The Optimist Daily: Making Solutions the N...

Published: Thu, 26 Mar 2026 00:00:13 +0000

Read more

← back to index