The Keywords Times

A personalized newspaper, written just for you.

Inspiration

In a world flooded with headlines, notifications, and algorithmic feeds, most news isn’t actually for us. It’s generic, reactive, and disconnected from our real lives. We scroll endlessly, yet still miss the news that truly matters—upcoming events, opportunities, deadlines, or changes that directly affect us.

That’s why we built The Keywords Times to be different.

Instead of chasing clicks, it listens to your real-world signals—your calendar, inbox, subscriptions, interests, and location—and synthesizes them into a daily, personalized newspaper. Every article explains why it matters to you, is backed by deep research and citations, and suggests a clear action you can take next.

The Problem

Today’s news ecosystem is optimized for profit, not interest.
Algorithms reward polarization and rage-bait because attention converts better than understanding. The result is negativity, disengagement, and passive consumption—news without context, agency, or follow-through.

Modern news media isn’t personal. And because of that, it often isn’t useful.

Our Solution

We created The Keywords Times—an AI journalist, just for you.

A newspaper that understands your life and turns information into direction.

What It Does

A Newspaper That Understands Your Life

The Keywords Times securely connects to your personal signals through MCP servers, including:

Email
Calendar
Social platforms
Content subscriptions

By analyzing this data alongside your background (location, school, job, interests), the system builds a living understanding of what’s relevant to you right now.

From upcoming hackathons to local events, weather shifts, tech trends, or career opportunities, stories are prioritized based on personal context, not engagement metrics.

Deep Research, Not Summaries

For every potential topic, an AI research agent:

Performs multi-source web searches
Verifies information across trusted outlets
Reduces bias and hallucinations

Each article includes citations and source attribution, so users always know where information comes from and why it’s trustworthy.

Action-Oriented News

Every article includes an Action Box—an AI-generated suggestion tailored to both the story and the user.

With one click, users can take action directly from the news.

Examples include:

Drafting an email or message
Adding an event to your calendar
Linking to applications or resources
Shopping suggestions driven by weather or timing

News doesn’t just inform—it moves you forward.

Personalized Games & Interaction

To make the experience engaging and playful, The Keywords Times generates:

Personalized Crosswords
Connections-style games

All themed around topics you actually care about.

How We Built It

The Keywords Times is a web-based application built with a modern, scalable stack.

Supabase securely stores user data and preferences
Keywords AI serves as the core orchestration layer for:
- Topic discovery
- Deep research
- Citation-aware article generation
- Personalized action suggestions
MCP servers (via providers like Composio) enable secure access to:
- Gmail
- Google Calendar
- X
- YouTube
- And more

To build and iterate rapidly, we used TRAE, an AI coding IDE, leveraging both Builder Mode and Solo Mode. TRAE allowed us to quickly scaffold complex agent pipelines and frontend components.

News and web search APIs power discovery, while a secondary deep-research agent verifies information across multiple sources before articles are written.

The frontend mirrors a clean, New York Times–style layout, emphasizing readability, hierarchy, and editorial polish.

Challenges We Ran Into

Fusing Personal Signals

One of the hardest problems was combining emails, calendars, social activity, subscriptions, and location into a single coherent model of “what matters right now.” Avoiding overfitting to one source while staying genuinely personal required careful agent design and prioritization.

Avoiding Bias Amplification

Generating personalized news without reinforcing bias is non-trivial. We designed a multi-step research pipeline that intentionally:

Seeks multiple perspectives
Validates claims
Filters out weak or sensational sources

All before article generation begins.

Making Actions Feel Natural

Most news systems stop at summarization. Designing reliable, useful action suggestions that felt natural—not gimmicky—required tight coupling between research, personalization, and reasoning. Every action needed a clear justification tied to that specific user.

Accomplishments We’re Proud Of

A Truly Personalized Newspaper

The Keywords Times doesn’t feel like a feed or a recommender system. It feels like a real newspaper—written for exactly one reader. Every article explicitly explains why it’s relevant, turning personalization from a hidden algorithm into a transparent feature.

Trust-First Journalism

By enforcing deep research and citation before generation, we built a system that prioritizes trust and clarity over speed. This dramatically reduces hallucinations and makes the output feel editorial rather than synthetic.

The Action Box

The Action Box is a core innovation we’re proud of. News rarely helps users move forward—we changed that by making every story end with a meaningful, contextual next step.

What We Learned

Personalization isn’t about more data.
The hardest part wasn’t accessing user signals—it was turning them into context-aware, respectful, and useful insights.

On the AI side, we gained deep experience orchestrating multi-agent workflows with research, verification, writing, and reasoning stages. On the product side, we learned how much users value clarity around why something is shown to them.