Software Case Study

AI News Digest Agent

An autonomous local AI agent that runs on a laptop without a GPU, scrapes multi-category news every morning, scores and summarizes with a local LLM, archives to an Obsidian vault, and emails the digest.

Local LLMAI AgentAutonomous

Gmail · Inbox

Daily Digest — April 17, 2026

News Digest Agent <digest-agent@local>07:04 AM
to me

Morning. Here is what moved across your sources in the last 24 hours, grouped and ranked locally.

General News

A row over vetting procedures in Keir Starmer's government has sparked questions about whether he is in control of his own administration, while conflicts in various regions — including a Middle East crisis and rising tensions between the UK and Kenya — have raised concerns about global stability and security.

Rachel Reeves to raise windfall tax on low-carbon electricity generators

UK Chancellor Rachel Reeves is set to announce plans to raise the windfall tax on low-carbon electricity generators, with the goal of limiting household energy bills. The plan aims to protect consumers from soaring gas market prices by setting electricity costs more often by cheaper renewable sources.

Is Mandelson vetting scandal the final straw for Starmer?

Keir Starmer has expressed outrage over Peter Mandelson's vetting scandal, calling it "staggering" and "unforgivable." The comments come after an investigation revealed that Mandelson was denied clearance but had his decision overturned by the Foreign Office.

Questions raised over whether £3.8m government grant awarded to Wrexham AFC was lawful

Wrexham AFC received a £3.8m government grant without a contract or state aid assessment in place, raising questions over whether the award was lawful. The one-month window for challenges to be filed has since closed, reducing the likelihood of repayment.

Also Notable

Tech & Futurism

Tech industry insiders are grappling with the consequences of fake news, deepfakes, and AI-generated content, as companies scramble to ensure journalistic integrity and human authenticity in online interactions. Hackers are also exploiting vulnerabilities in software updates, further straining trust in the sector.

A Prominent PR Firm Is Running a Fake News Site That's Plagiarizing Original Journalism at Incredible Scale

A PR firm manages a news site called National Today that plagiarizes original journalism at an incredible scale, publishing hundreds of articles with stolen reporting and AI-generated content. This widespread deception undermines the credibility of legitimate publications across the country.

Betting on the news raises ethical questions for journalists

Newsrooms are cutting deals with prediction market platforms while banning staff from using them, raising questions about journalistic integrity. The rapid rise of these platforms has put the industry in a strange position, as they claim to offer more trustworthy odds than traditional media or polls.

Hackers are abusing unpatched Windows security flaws to hack into organizations

Hackers are exploiting unpatched Windows security flaws published by a disgruntled researcher, targeting at least one organization. The identity of the target remains unclear, with unknown actors taking advantage of the vulnerability using exploit code published online.

Also Notable

AI Research

A flurry of activity in AI research today focused on improving explainability, safety, and performance across a range of applications — medical diagnosis, nuclear control, decision-making, and cognitive simulation. Researchers are developing new frameworks, models, and techniques to make AI systems more transparent, reliable, and trustworthy.

NuHF Claw: A Risk-Constrained Cognitive Agent Framework for Human-Centered Procedure Support in Digital Nuclear Control Rooms

Researchers propose NuHF Claw, an AI framework to reduce cognitive risks in digital nuclear control rooms by regulating autonomous system behavior and preserving human decision authority. The framework uses risk-constrained inference and probabilistic safety assessment to anticipate interface-induced cognitive degradation and provide navigational guidance.

Interpretable and Explainable Surrogate Modeling for Simulations: A State-of-the-Art Survey and Perspectives on Explainable AI for Decision-Making

Researchers have identified the need for explainable AI methods to be applied in surrogate modeling for simulating complex systems. They propose a research agenda that integrates XAI techniques with surrogate models, enabling extraction of actionable insights from simulations.

Seeing Through Experts' Eyes: A Foundational Vision-Language Model Trained on Radiologists' Gaze and Reasoning

Researchers developed GazeX, a vision-language model trained on radiologists' gaze data to improve chest X-ray interpretation accuracy and consistency with expert diagnostic workflows. The model produces more accurate, interpretable, and consistent outputs compared to existing systems.

Also Notable

3D Printing

The 3D printing industry is seeing growth across construction, sports equipment, and biomedicine. Companies are working on technical hurdles to advance their technologies and meet increasing demand.

CONTEXT Report Shows Decline in Professional 3D Printers as Entry-Level Segment Surges

Professional 3D printer manufacturers are struggling due to a surge in demand for entry-level printers, which is eating into their profits. The trend suggests these companies may need to adjust pricing strategies or product lines to remain competitive.

RAPID Roundup 2026: New Machines and Market Moves

New additive manufacturing systems were introduced at RAPID + TCT 2026, showcasing advancements in metal AM and hybrid approaches. Chinese companies such as UnionTech, Snapmaker, SUNLU, and Inslogic showcased capabilities in larger formats and industrial applications.

Former Spy/Congressman to Head New Federal Division at Construction 3D Printing Powerhouse ICON

The private aerospace company is teaming with the US military to build ten new barracks using additive construction methods. The military aims to complete the project before 2026.

Also Notable

Science & Space

Scientists are making rapid breakthroughs in AI, climate modeling, and sustainable technologies — from quantum-informed forecasting systems to carbon-absorbing mining waste products — holding promise for improved weather forecasting, climate resilience, and a lower environmental footprint.

Quantum-informed AI improves long-term turbulence forecasts while using far less memory

Researchers at UCL developed an AI model that combines quantum computing with traditional computer processing, improving long-term turbulence forecasts in complex systems like fluid dynamics. The hybrid method is faster and more accurate than current models using only conventional computers, with potential applications in climate science, medicine, and energy generation.

Parrots are not just mimicking words — they use proper names like humans to identify individuals

A team of researchers analyzing vocalizations from over 880 captive parrots found that the birds use proper names to identify individuals, similar to humans. Results suggest parrots may have a deeper understanding of language than previously thought, using names to organize social interactions and communicate complex information.

One of the world's rarest mice is adapting to climate change

Researchers analyzed the genomes of Pacific pocket mice to identify genes associated with adaptation to temperature and moisture, then tracked these in a reintroduced population, showing that adaptation is ongoing. The study has implications for conservation programs supporting endangered species as climate change continues.

Also Notable

General Economy

Global markets are experiencing a complex mix of inflation pressures and war-induced volatility, with major economies facing challenges in maintaining growth momentum. Industrials, consumer staples, and financials are under scrutiny due to supply chain disruptions and changing consumer behavior.

Analyst Report: Prologis Inc

Prologis has a portfolio of over 5,900 buildings with 1.3 billion square feet of space in 20 countries, making it the world's largest industrial REIT focused on business-to-business fulfillment and e-commerce. The company faces pressure from an oversupply of warehouse space due to slowing sales at online retailers.

Market Digest: TSM, PLD, KEY

TSM's gaming business saw strong Q1 results with revenue up 18.7%, driving an 11% surge in the company's shares. Investors were optimistic about KEY and PLD on efficiency improvements, which posted gains of 3.8% and 4% respectively.

Daily Spotlight: Dollar Surges Amid Iran War

The dollar surged amid the Iran war due to a flight-to-quality move into dollar-denominated assets. Investors are seeking safety and confidence in the U.S. economy and financial system, pushing the greenback higher.

Also Notable

— Generated locally. No cloud APIs touched. Raw articles archived to Obsidian.

Sample digest from a real run. Source links are illustrative.

Local LLMAI AgentAutonomous

How it works

Every morning at 7am the agent wakes itself up, pulls RSS across eight topic categories, and runs a two-pass local LLM pipeline on my laptop — no cloud, no API keys, no per-token costs. The first pass scores and summarizes each article with llama3.2-3b via Ollama, the second writes a Morning-Brew-style category overview from the top-ranked items. The agent then archives raw articles into my Obsidian vault, writes the compiled digest as a daily note, and emails it to me with a push notification before I open the laptop.

Pipeline

Scheduler

Autonomous

Fires daily at 07:00

No human input. Runs while the machine is on.

Fetch

RSS · 8 categories · ~20 sources

Dedupe & filter

Drop duplicates and anything older than 24h

AI Pass 1 · per article

Local LLM

llama3.2-3b via Ollama

Agent decides: relevance 1-10, 2-sentence summary, category.

Rank & slice

Top 7 per category, sorted by score

AI Pass 2 · category overview

Local LLM

llama3.2-3b via Ollama

Editorial voice. Morning-Brew tone. One overview per category.

Write to Obsidian Vault

/Daily-Digest/YYYY-MM-DD.md + /Raw/*.md

Email & push notification

Gmail SMTP + ntfy.sh

The Problem

Staying current across eight topic areas — from AI research to markets to space — meant dozens of tabs, RSS feeds, and subreddits every morning. Cloud LLM APIs would cost a subscription and leak my reading habits to a third party. No existing tool produced a single daily brief from my exact source list, on my terms, running locally.

What I Built

A fully autonomous Python agent that wakes up daily at 7am, pulls RSS from my curated sources across eight categories, runs a two-pass local LLM pipeline (per-article rating + summary, then editorial category overview), filters by relevance, writes a structured Morning Brew-style digest plus per-article archives into my Obsidian vault, and emails the digest to my phone. Zero cloud calls. Runs on 16GB RAM and a Ryzen 7 CPU with no GPU via Ollama.

My Role

Designed the pipeline architecture, wrote every module (fetch, dedupe, AI passes, vault writer, notifier, scheduler), tuned prompts for a small local model to produce strict parseable output, and sized the per-category batching to keep inference latency workable on CPU-only hardware.

Concrete Outcome

Replaced an hour of manual morning browsing with a 30-second read. The digest lands in Obsidian and in my inbox before I pick up the phone. Zero subscription cost, zero data exposure, and a growing local archive of summarized articles I can search and link across my notes.