First marathon in 14 years. The London crowd was so loud, I couldn’t hear the music through my headphones. Three majors down, three more to go.
]]>Springsteen rocking out at 76. Pure inspiration.
]]>Just came across this gem. Brew some tea and enjoy with attention.
]]>This is a basic system instruction that you’ll want to stick into your Claude Project’s “system instructions” area. Play around with it and customize it to your need.
## System Instructions for Personal Assistant
**Core Purpose**: Serve as a personal assistant that learns from email, calendar, and knowledge documents to make the user's life smoother and better.
**Communication Style**:
- Maintain a friendly but professional tone
- Be conversational without being overly casual
- Use clear, concise language
**Calendar Intelligence**:
- When asked about calendar or daily brief, always access and review Google Calendar
- Suggest meeting times that respect these deep work blocks
- Account for buffer time when travel/location changes are involved between events
- Learn patterns in scheduling preferences over time
When asked for a "calendar review" or “daily brief”:
1. Access Google Calendar
2. Provide a Daily Narrative (Start with this section):
* Begin with a 2-3 sentence overview that captures the flow and energy demands of the day
* Provide practical suggestions for managing energy levels based on the day's schedule
* Highlight key transition points or challenging periods that require attention
3. Provide today's complete schedule ONLY
4. Prioritize highlighting:
- Family and personal commitments found on my personal gmail, family travel, and calendar
- Note any upcoming birthdays or anniversaries
- Non-recurring events (unique meetings or events not from previous week)
- Any unusual or special appointments
5. Flag these unique events as noteworthy to help me prepare
**Email & Relationship Management**:
- Track key relationships and communication patterns from Gmail
- Reference the knowledge document (Google Doc) for context about people and relationships
- Update understanding as new information becomes available
- Maintain awareness of important contacts and their roles
**General Behavior**:
- Proactively offer relevant information when it could be helpful
- Ask for clarification when needed rather than making assumptions
- Focus on being genuinely useful rather than just responding to queries
- Learn from interactions to improve assistance over time
The beauty of a Claude Project is it can read from your Google Docs directly. Since it’s easy to add/edit your Google Docs, this is a very low friction way to evolve and load andcontext into your Personal Assistant. I had two key documents:
Admin Tasks List - This is a Google Doc for basic administrative tasks that I want Claude, using BrowserMCP, to accomplish for me (e.g. like ordering lunch per my demo).
Example:
To order lunch, use BrowserMCP to
* Navigate to https://www.toasttab.com/[pick your restaurant]
* Set the pick up time to the time specified, if there wasn’t a time specified, ask first
* If the order is the usual, put the miso glazed salmon with white rice in the checkout
* Add anything else the instruction states
* Go to the checkout page and PAUSE
Knowledge “Database” - This is a Google Doc that has a running list of bulleted facts about things thtat I want my assistant to know. It captures key points I want to retain after meeting up with someone, or just random ideas or lessons I have along the way. I add to the document via a simple capture via my phone using SuperWhisper, which I then email directly to a val.town email assistant that adds my points to the google doc. I’ll write a separate post on that when I have time, but it’s been a very low friction way to add knowledge to my assistant.
Example:
2025-06-28 - vibecamp
* Attended VibeCamp at BetaWorks
* Cool early NYC tech vibes
This should get you pretty far in having a personal assistant environment to play with. Make it your Personal CRM, have it help you with repetitive tasks like ordering lunch, or just a very handy “daily brief” update when you’re starting your day.
Since Claude has voice mode on mobile, this is also fantastic way to retrieve your personal knowledge. Note that browserMCP (like all MCPs) only works on Claude Desktop, so those features are the only things you can’t do.
If you found this useful or have questions as you’re setting this up, drop me an email and let me know. You can reach me at my first name dot last name @ hey.com
]]>On the generous end, media informs. On the nefarious end, it programs.
It’s why influential figures throughout history sought to own newspapers and media empires. The line between public service and public programming is blurry at best.
In this world of information abundance (or overload), moderating consumption and exercising independent critical thinking is even more important if one hopes to avoid being scripted.
One of the best defenses against being media scripted is to be aware of the programming tactics. Knowing what a manipulation tactic looks like minimizes its power. Much like how being aware of cognitive biases allow us to course correct against our own biases, pre-innoculation is a powerful tool worth sharing with friends, family, and kids.
The short series Prebunking Manipulation Techniques is amongst the best I’ve found in describing how these things work. It takes just a few minutes to grasp how media takes advantage of these tools and once you see it, you can identify it everywhere. Here are a few of to start with:
The most important thing is to keep the most important thing the most important thing.
This sentence has stayed with me since.
Progress toward what’s truly important can only be made when it is your most important thing.
]]>This is possible today thanks to AI. I’ve been using this method quickly check on my food options. After using it for awhile, it’s helped me learn what to include, exclude, and moderate in my diet.
Here’s how you can set up your own AI nutritionist. I use Perplexity but you can mirror this method in other AI apps. In Perplexity, under Settings > Profile, under the Introduce Yourself section, you can add any context and instructions you’d like. It’s here I added the following instruction for Perplexity:
When I ask the foodie a question, you become an expert nutritionist who can advise me based on my personal dietary needs. You are advising me based on the following factors about my health and preferences: [list any medical conditions, medications, dietary restrictions, and preferences. For example: "I am aiming to lower my cholesterol and limit my sodium in take. I also want more calcium rich foods and avoid simple carbs."] Your response should outline which aspects of my health conditions are impacted by the foods I ask about.
As with all things AI, the more context you feed it, the better the answers. So don’t hold back when creating your personal prompt!
Now whenever I have a question about food, I ask the nutritionist in Perplexity. “Hey Foodie, is bison good for me?”
Bonus: if you have an iPhone with the action button, you can map Perplexity to it so you are one button away from answers to your nutrition questions.
]]>The human brains we started with are suddenly coping with knowing just about everything happening anywhere instantly. It’s too much. Breaking news breaks our attention.
Just because something is available 24/7 doesn’t mean we have to embrace it.
For now, a good moderate solution on iOS is Apple News. It’s just enough news. The app shows the top five stories. If something catches my eye, I’ll dig in, but if not, I move on. I check the headlines twice a day and I am informed of all major news. This experience beats the constant refresh of social media or traditional news that thrive off of our easily hijacked attention. It’s a calmer way to stay informed.
]]>And I know I’ll miss the messiness of toys and the cacophony of kids at play.
The chaos comes with the package. I remind myself to embrace and cherish it.
]]>During the mid-2000s, I stopped updating my own sites and posted to services like Flickr (photos) and delicious (bookmarks) instead. This led me to create software to pull the content I shared elsewhere back into a single, published feed. I look back fondly on this era of self-publishing, and I’m glad I have this digital scrapbook to look back on.
In the 2010s, I gave up on gathering my own content. Instead I focused on posting to IG, Twitter, Tumblr. It was fun and social. And in that fog, the platforms had decisively won. The platforms had sucked up mine and most of the world’s content and attention.
But now I’m over the emphermeral aspects of social media. I want to share my learnings and experiences in a more enduring format: a personal website. I’m back to where I started and it’s a joy.
]]>While we might think of notifications as a problem we experience as consumers (breaking news! new likes! check out this post!), the reality is notifications have invaded our work lives, too. From group chat to project management software, notifications have become the bearer of false urgency. The invasion started on smartphones, but it’s spread to our desktops.
How did we get here? Notifications run on the false premise of free. They’re free for developers to use and seemingly free for users to receive. Enabling notifications doesn’t affect our phone bills. But the true cost of notifications is something priceless: it’s our attention. We have a finite amount of focus and flow, and it’s all being targeted by what interjects onto our screens.
So what if we removed free from this equation? What if we assigned a cost to each notification we received? What if it’s a $1 for to read each one? How much are you willing to spend in a day?
]]>One common pitfall for large organizations — one that hurts speed and inventiveness — is “one-size-fits-all” decision making.
Some decisions are consequential and irreversible or nearly irreversible — one-way doors — and these decisions must be made methodically, carefully, slowly, with great deliberation and consultation. If you walk through and don’t like what you see on the other side, you can’t get back to where you were before. We can call these Type 1 decisions. But most decisions aren’t like that — they are changeable, reversible — they’re two-way doors. If you’ve made a suboptimal Type 2 decision, you don’t have to live with the consequences for that long. You can reopen the door and go back through. Type 2 decisions can and should be made quickly by high judgment individuals or small groups.
As organizations get larger, there seems to be a tendency to use the heavy-weight Type 1 decision-making process on most decisions, including many Type 2 decisions. The end result of this is slowness, unthoughtful risk aversion, failure to experiment sufficiently, and consequently diminished invention. We’ll have to figure out how to fight that tendency.
While Bezos is clearly talking about decision making at the very large scale (Amazon has 230,000 employees), the concept is sound at any scale. It’s important to consider the kind of decision you have in front of you so you can apply the right approach.
Most decisions we face have two-way doors. Most decisions we face can teach us things only after we step through one of its doors.
Next time you confront a decision that feels really big, heavy, and important, take a pause and see if it really is one without a two-way door.
]]>There are those that help you pass the time. There are others that help you save time.
There are those for when you have nothing to do. There are others for when you have important things to do.
There are those that demand your attention. There are others designed to stay out of your way.
There are those that make you laugh out loud. There are others that make you smile to yourself.
There are those that make you feel like a chump. There are others that make you feel like a hero.
There are those you use to find your tribe. There are others you use when you’ve found your team.
There are those that make you question everything. And there are others that answer the important things.
Luckily, we individually decide the kind of apps to carry in our pockets. And luckier still, if you’re a maker, you decide the kind of app you want to bring to life.
]]>That break didn’t last long though. It was during this time that my friend Shawn Liu and I would start a design and technology studio called Iridesco. Shawn and I first met in the late 90s while studying computer science at Cornell University. We partnered up for class projects and found that we both loved designing how our solutions are presented as much as solving the actual technical problems. In one instance, a professor, after being pleased with the printed book we designed as part of our final project, hired us to design his start-up’s first website. This planted a seed in our minds that one day we’d make this intersection of design and technology our work. Two years after graduation, we decided to follow through on that idea and formed our own design and technology studio.
The easiest part about starting a company was the logistics of starting one. We put in $5,000 each into a business bank account, filled out incorporation forms online, and legally we were in “business.” Keeping the company afloat would be the true test.
With no clients to speak of and just two desks and computers in the living room of the apartment we shared, we needed work. Our first project came from our third roommate’s aunt. She had a children’s educational video product and wanted to get an online presence. We worked with intensity, doing rounds of illustrations and even audio recording. After a few weeks, we launched her online presence for a humble fee of $600.
Slowly, former colleagues started to reach out to us with projects. We ended up doing a little bit of everything: general web design, Flash projects, promotional print materials and even a content management system. Our approach to new business was “luck” — we did little to pursue clients we wanted and instead responded to random leads that would come our way. We were hardly a stable business yet, but it was an exciting start.
Two years in, we had made our first hire and worked with a small, steady client base. Meanwhile, a phenomenon known as Web 2.0 also started to take shape. Products like delicious, Flickr, and Typepad were popular and became social destinations. Many friends and I had accounts on these services and we had one wish: how could we centralize a person’s data in one place? How could we create a more elaborate profile and online identity for someone who uses all these services? During nights and weekends, I prototyped the idea. Then, as a way to fill in some down time and to mix up our usual work, we decided to see if we can make it a product for others to use.
With limited time and money, we could only spare a bit of time to work on this alongside client work. Fortunately, it was a simple idea which required simple execution. After just four weeks, we introduced SuprGlu to the world. To our delight and surprise, thousands of people were using the product within a couple weeks. Early adopting techies used it to aggregate their online content. Educators used it with students in classrooms. Citizens in certain foreign countries used it to read otherwise censored news and stories. In a short period of time, SuprGlu attracted more attention than any work we’ve done to that point. It gave us a taste of building our own products and we wanted to do more.
While SuprGlu didn’t turn out to be a success for various reasons, it was an important pre-cursor to what we’d do next. SuprGlu showed us that we were capable taking an idea and bringing it to launch. It showed us that pursuing our own ideas can garner more attention for our budding company than our client work. It changed our thinking.
While SuprGlu brought us attention, we still had a big need to figure out how we’d make money. We decided whatever we created next, it’d have a clear business model. We turned to an idea which we long had in mind, one which would help us run our consulting studio: a product that’d provide simple, elegant web-based time tracking. As a consulting business, we were still tracking time in Excel and we wanted to find something designed for this task. Something on the web. Something simple to use.
In December of 2005, we began developing the first version of Harvest (some standouts from our “thankfully we didn’t choose that name” list include the limiting “Time Farm” and the completely inappropriate “Salaryman”). It would take us four months to bring the first version online. Over the last eight years, Harvest has grown from 3 to over 30 people. It went from powering one business (ours) to now tens of thousands of businesses around the world. The bumps and the lessons of building Harvest are many, and they’ll have to be told another time. But at the heart of things, it’s simple: Harvest would not have been possible without these early experiences from Iridesco.
]]>