FundamentalsExplainer10 min
Fine-Tuning-as-a-Service: What It Is and What It's Used For
How hosted fine-tuning works under the hood, what organisations use it for, and why a few training examples can undo a model's safety.
LLMs
The workhorse of the field, and the thing most coverage describes without explaining. What I care about is the mechanism underneath a behaviour, not the demo that showed it off.
Fundamentals5
FundamentalsDeep dive9 min
From Experts to Sub-Experts: Smarter Fine-Tuning for MoE Models
NSFT cuts each expert in a Mixture-of-Experts model into channel groups and trains only the ones that matter — beating LoRA and ESFT on accuracy per trained parameter, and forgetting less along the way.
FundamentalsDeep dive5 min
Why One Pruned Model Can't Do Everything: Task-Aware Spectral Pruning
Classic LLM pruning commits to one compressed structure before it knows what question a user will ask. TASP calibrates five task-specific masks instead, and a cheap router picks one per turn.
FundamentalsExplainer18 min
Tokens: the one idea that explains most of AI's odd behaviour
Why a model that can explain quantum tunnelling cannot count the letters in strawberry, why long conversations get slow and expensive, and why Dutch costs more than English. It all comes back to the very first thing that happens to your text.
FundamentalsNews2 min
Understanding Tokens in AI Language Models
A plain-language guide to what tokens are and why they matter in AI language processing.
Scout
Research11
ResearchPaper5 min
OmniEdu: Revolutionizing Education with Open Foundation Models
Exploring how OmniEdu's open foundation models enhance K-12 learning through improved problem-solving, curriculum grounding, and instructional support.
ResearchPaper5 min
When AI Reviews Train AI Reviewers: Scientific-Judgment Collapse and Mitigation
Exploring how AI-generated reviews can lead to repetitive evaluation patterns and discussing methods to maintain diversity and accuracy in AI-assisted peer reviews.
ResearchNews5 min
Claude Code v2.1.267: Enhancements and Fixes You Need to Know
An in-depth look at the updates in Claude Code's latest release and what they mean for developers.
ResearchNews5 min
Navigating the 0.155.0-alpha.3.8 Release: What's New in OpenAI Codex
An in-depth look at the latest updates and what they mean for developers using the OpenAI Codex.
ResearchDeep dive9 min
How Do You Catch a Machine in a Lie?
Nineteen teams spent a month building AI lie detectors. The winning trick turned out to be less clever than it looked — and that's the interesting part.
Sebastiaan with AI
ResearchPaper5 min
Skills or subagents? How you run reusable knowledge matters as much as what's in it
A new paper compares two ways of executing agent skill packages, and finds that subagents only win when a skill declares what it needs and what it returns.
ResearchPaper7 min
LLMs Model a Harsher World Than the One We Live In
A new benchmark shows language models know what's wrong but not how people actually react to it — and the bias may come from alignment itself.
Sebastiaan with AI
ResearchNews5 min
Building an AI That Learns From Experts: Creating an Organizational Second Brain
How Meta's AI agent captures and preserves expert knowledge, turning fleeting insights into durable institutional memory.
ResearchNews5 min
Exploring NVIDIA's Personal AI Router: A New Era of Distributed AI Computing
Understanding NVIDIA's open-source PAIR and its potential to transform local AI processing.
ResearchNews5 min
Inside Microsoft's Project Zenith: A New Era for Developer-Focused Windows
Exploring how Project Zenith reshapes the Windows experience for developers with its focused, distraction-free setup.
ResearchNews1 min
Nvidia's Acquisition of Hugging Face: What's at Stake for AI Development
Analyzing the implications of Nvidia's $12.9 billion purchase of Hugging Face.
Scout
Benchmarks3
BenchmarksBenchmark1 min
New Leaders Emerge in AI Benchmarking
Recent benchmark results reveal new top performers and first-time measurements in AI model evaluations.
Scout
BenchmarksBenchmark1 min
Discrepancies in LLM Performance Claims: A Closer Look at GPQA Diamond
Vendor claims for glm-4-7-flash significantly exceed independent measurements.
Scout
BenchmarksBenchmark1 min
Benchmark Showdown: GPT-6-Astra and GPT-5-5 Lead the Pack
New leaders emerge in the AI benchmark arena as GPT-6-Astra and GPT-5-5 set new standards.
Scout
Security4
SecurityDeep dive8 min
GLM-5.3: Frontier-Grade Cyber Capability, Released With the Locks Left Open
Z.ai says GLM-5.3 is a coding model ready for cyber defense. Anthropic says it builds exploits like its own restricted model, but with safeguards that barely hold. A weighed look at both claims — and what it means in practice.
Sebastiaan with AI
SecurityNews5 min
Understanding Adversarial Attacks on Large Language Models
Exploring the vulnerabilities in AI models and the implications for enterprise security.
Scout
SecurityDeep dive17 min
Abliteration: how one direction holds a model's refusals
Open-weight models ship with guardrails. A technique borrowed from interpretability research removes them in minutes, with no retraining and no prompt trickery. What it actually does, how far it goes, and whether anything can be done about it.
Sebastiaan with AI
SecurityNews4 min
Unpacking Abliteration.ai: The Business of Removing AI Guardrails
Exploring the implications and applications of Abliteration.ai's unguarded models in cybersecurity and beyond.
Scout
Enterprise AI1
Enterprise AIDeep dive11 min
Claude Duet: what it is, and what it can't do yet
Anthropic's new proactive agent watches your apps, works while you're away and asks before it acts. A practitioner's look at what it does, where it falls short, and what it means for IT governance.
Sebastiaan with AI
Model Updates20
Mistral Large 4 (Le Chonk) Released with Open Weights
Mistral has released Mistral Large 4, a 1-trillion-parameter multimodal model, now available with open weights.
OpenAI Launches GPT-6.1 Terra
OpenAI releases GPT-6.1 Terra, enhancing speed and efficiency for enterprise tasks.
Terminal X Enhances Investment Workflows with GPT-6.1 Sol
Terminal X has launched GPT-6.1 Sol, significantly improving investment judgment accuracy and efficiency.
Microsoft Releases Sentiment-Analysis-7B Model
Microsoft has launched Sentiment-Analysis-7B, a 7B-parameter language model for sentiment classification.
NVIDIA Releases GLM-5.3-Flash-FP4-Local Model
NVIDIA has launched the GLM-5.3-Flash-FP4-Local model, a quantized version of ZAI's GLM-5.3-Flash model, optimized for local AI hardware.
Google Releases Gemini 4 Argon for Complex Tasks
Google has launched Gemini 4 Argon, their most advanced model for complex coding and scientific tasks.
Liquid AI Releases LFM2.5-24B-A2B: A New Reasoning Model
Liquid AI launches LFM2.5-24B-A2B, a 24B-parameter reasoning model designed for fast, private on-device agents.
Liquid AI Launches LFM2.5-24B-A2B-Base Model
Liquid AI introduces the LFM2.5-24B-A2B-Base, a 24B-parameter hybrid model optimized for on-device deployment.
Claude Haiku 5.5 Launches on AWS
AWS introduces Claude Haiku 5.5, enhancing speed and efficiency for subagents and high-volume tasks.
GLM 5.3 by Z.ai Now Available on Amazon Bedrock
Z.ai's GLM 5.3 is now generally available on Amazon Bedrock, offering enhanced capabilities for agentic coding and software engineering.
OpenAI DevDay 2026: the new features at a glance
Dots, GPT-6.1 Sol, Ultrafast, the Agents API and plugins — the announcements that matter from DevDay 2026, and what they mean if you administer IT.
Empowering Enterprises with ChatGPT's New Data Agent
How OpenAI's Data Agent in ChatGPT Work transforms enterprise data management with its intuitive and dynamic approach.
Gemini 3.7 Flash: The New Frontier in AI Workhorse Models
Unpacking the improvements and promises of Google's latest AI model for coding and agent workflows.
Google's August AI Announcements: A Closer Look at Gemini and Pixel 11
Exploring Google's latest AI advancements with the Gemini models and Pixel 11 devices.
GPT-6 Astra: A New Era for Developers
Explore the capabilities of GPT-6 Astra, a revolutionary AI model that's reshaping how developers approach complex projects, from 3D modeling to intricate design tasks.
GPT-6 Astra: Redefining Intelligence and Efficiency for Developers
Exploring the leaps in capability and alignment with OpenAI's latest model, GPT-6 Astra, and how it stands against competitors like Anthropic's Fable.
Navigating the Pricing Maze with Anthropic's Claude Commerce Agents
Exploring how Anthropic's new commerce agents can transform pricing strategies for retail and beyond.
Understanding the Impact of OpenAI's GPT-6 Astra on Enterprise Operations
An exploration of OpenAI's new GPT-6 Astra model, its potential costs, performance implications, and expected applications in enterprise environments.
NeoMME: A Breakthrough in Multimodal and Multilingual Encoding
Exploring how NeoMME's efficient design reshapes visual document retrieval.
Anthropic ships a faster, cheaper reasoning model
What the new release changes for everyday agent workloads.