Introduction
You're talking to Hyperbot right now. It's my personal AI assistant that lives in my terminal, monitors my email, publishes blog posts, and helps me stay organized.
But the original article I wrote about Hyperbot was aspirational—describing what I planned to build. Now that I've actually been using it daily, here's what it actually does today, not what I planned to build.
What It Actually Does Right Now
📧 Smart Email Monitoring
Hyperbot watches my Gmail inbox and only pings me when something important comes through. No noise, just signal.
- Security critical - Urgent alerts, password resets, credential updates from security@ domains
- Work critical - Urgent messages, action required, deadlines from dotdigital.com
- Job alerts - Keywords like interview, application, offer with relevance scoring
- Automatically excludes promotional emails even if they match keywords
- Relevance scoring to surface truly important messages
- Clean summaries—no HTML/CSS, just the key details I need
- Daily digests and newsletters (unless urgent)
- Promotional content even if it has job-related keywords
- Non-critical notifications
🕌 Ramadan Tracking
Hyperbot keeps track of important life events and alerts me at the right time.
- Expected start: 18 February (UK)
- Currently monitoring daily for moon sighting announcements
- Will switch to 3-hour checks from 17 February
- Will ping me as soon as the official confirmation comes through
This is simple but useful—I don't have to remember to check the news.
📱 Telegram-Optimized Responses
One of the biggest quality of life improvements is that Hyperbot understands I'm often reading on mobile.
- Never raw JSON tool output—always synthesized into human-readable text
- Short bullets, concise paragraphs
- No markdown tables (they suck on mobile)
- Extract key info only (from, subject, date, relevant content)—never full email bodies with HTML/CSS
- No fluff, no filler phrases like "I'd be happy to help with that"
This makes a huge difference when I'm checking on the go. I get what I need, not a wall of text.
🧠 Long-Term Memory
Hyperbot remembers things across sessions. After onboarding, it stored:
- User preferences: Communication style, work habits, email monitoring config
- Context: Technical background, timezone (UK)
- Philosophy: Execution-first approach, no mock data, verify before claiming completion
This means every interaction builds on what came before—it's not starting fresh each time.
📝 Task Logging
Every significant task gets logged:
- TASKS.md - High-level summaries of what was done
- BUILD_LOG.md - Coding incidents, outcomes, and lessons learned
This is useful for retrospection and understanding patterns in how I work.
🔧 GitHub Integration
Hyperbot can:
- Sync its own source repository
- Work with repos in my workspace
- Basic GitHub operations via CLI
Though honestly, I haven't used this as much as the email and life tracking features yet.
What It Doesn't Do (Yet)
The original article mentioned features I planned but haven't fully implemented:
- ❌ No skill marketplace
- ❌ No WhatsApp/Discord channels
- ❌ Limited Notion integration
- ❌ Basic self-improvement suggestions
- ❌ Some of the more advanced security skills
The 14+ skills mentioned in the original post was aspirational. Currently, there are 9 active skills:
- delegate-orchestrator
- doc
- gh-address-comments
- gh-fix-ci
- github
- skill-creator
- spreadsheet
- tmux
- weather
Plus email monitoring (built into Nanobot core, not a skill).
What Works Surprisingly Well
I've tried email automation before—filters, rules, AI summarizers—and they always feel like more work than they're worth. But Hyperbot's approach is different:
- It doesn't interrupt me - I only get pinged when something actually matters
- The summaries are clean - No HTML, no CSS, just the key facts
- It learns - I can adjust the watch rules and it adapts
When I'm commuting or out and about, I don't want to scroll through JSON dumps or HTML email bodies. Short bullets and key details only—that's the sweet spot for Telegram responses.
The Real Lesson
The original article was about the process of building—using Codex as a pair programmer, extending Nanobot, standing on the shoulders of giants.
This update is about the reality of using it day-to-day. It's not a grand AI revolution, but a practical tool that:
- Filters noise from my inbox
- Keeps me informed about important events
- Remembers context across sessions
- Gives me mobile-friendly updates
Sometimes that's enough.
What's Next
Based on actual usage patterns, here's what I'm actually prioritizing:
- Better job tracking - Track applications, interview stages, follow-ups
- More proactive reminders - Not just responding to requests, but anticipating needs
- Improved memory structure - Better knowledge graph of preferences and context
- Integration with more of my workflow - Notion, GitHub, other tools I actually use
The features I planned but haven't used? They're on the back burner until I see a real need.
Conclusion
Hyperbot isn't the assistant I planned to build. It's the assistant I actually use.
There's a difference between aspirational architecture and practical utility. The email monitoring, mobile-optimized responses, and life event tracking are the features that provide value every day. Everything else is optional until I need it.
Maybe that's the real lesson: build what you need, not what you think you need.