Ping a friend

[Me, 28-01-26 22:37] What kind of strings are on Feist's guitar in Forever Before?

[Ping] Send text to Garrett.

Ping is a search engine project that explores the tensions, value systems, and hierarchies of where we seek out information. In Ping, the search engine database holds my phone contacts and the knowledge areas that I attribute to each person (e.g., Anya: [dogs, gardening, farm, sewing, reality TV]). When I search for something on Ping, the result isn’t a list of websites or an answer but rather the name of who I should ask my question to.

Technological progression (especially search and chat-based AI systems surrounding information seeking) places efficiency as a top, guiding principle—the faster the path from question to answer the better. Google even ranks websites with a faster page load more highly in search results than those that load slowly (Google, 2008). In each step of technological progression from oral histories to written histories to libraries to search engines to AI chats, speed increasingly dominates, and sources become secondary confirmations of reliability rather than the entry points to information.

While there is certainly a gained efficiency and resulting autonomy through these technological advancements (we no longer “need anyone” to answer our questions), this double-edged sword also silently degrades a more foundational level of value that we place in our community of family, friends, and strangers.

What would it be like to once again rely directly on the knowledge of the people that surround us? What if efficiency to an answer wasn’t the top, guiding principle? Could the efficiency loss be outweighed by the interpersonal connection created, a two-way interaction rather than a vacuous one to a server? What if efficiency is a blinder that upholds our biases and stifles opportunities for spontaneous connection? 

As the dopamine-tapping nature of search engines and AI continue to pull for our attention and trust, I hoped to reveal to myself the ways I might overrely on a global network of servers for information and under-rely on the knowledge of my community.

[Garrett in reply, 28-01-26 22:40] Nylon for sure

But sounds crazy

U should look at silk and steel I think they r called

It’s like a hybrid

Sounds like that

[Me, 28-01-26 22:41] Yeah I was thinking it was nylon too but couldn’t quite tell, such pretty tone  

Songs amazing

Never heard it

We hold implicit value systems and hierarchies towards where we go for different types of information and how much we trust the answer we receive. It’s more complex than just I ask friends and family for emotional advice, I ask the internet how to blah blah blah, I ask the shop worker if they have an item back in stock. In reality, it’s a blend of emotional, cultural, situational, and knowledge-related factors that influence where we seek out information in a given moment. For example, “Eileen knows a lot about this topic, but I know she’s so busy with work right now so I won’t bother her.”

Yet it seems the immediacy of the internet is often winning as we increasingly rely on search engines over people as our always-on knowledge source, and it’s creating a compounding dependence. As the brain gets accustomed to having access to the internet, it is less likely to commit the information to memory because we know we could just search for the answer again in the future (Gong and Yang, 2024). If we ask the internet once, we’ll likely have to ask it again.

And the reality is that as search and chat-based algorithms have gotten more and more advanced, beyond knowledge-based information seeking, we are increasingly turning to them for advice and emotional connection.With the rise of ChatGPT, Character.ai, and other AI systems, a set of 2025 surveys revealed that 1 in 3 US teens used AI companions for social interactions and relationships and more than 1 in 4 adults claimed to have had at least one intimate or romantic relationship with AI (Forbes, 2025).

It’s not to say that these technologies aren’t helpful to people, but what do we lose between people, as a society, when we turn to a digital interface as our extended memory, our confidant, or even partner? The immediacy and unwavering availability of these technological systems can be a crutch that, when over-relied upon, removes opportunities for us to learn from and support each other.

[Me, 26-01-26 13:07] Can people eat the oranges that fall off public citrus trees in Barcelona?

[Ping] Send text to Madu.

In 1983, “ping” became the name of a network utility tool to measure the speed of a response from one computer to another (named after the sound sonar makes) (Wikipedia). Over time, we’ve evolved the use of the word to apply not just to computer to computer connections but also to our human interactions as well—“Ping me when you’re free to chat.” It seemed a fitting name for a search network made up of the people I know, and also a slightly alarming reflection on how, through language, we synonymize ourselves with computers.

I built Ping to look like a standard search browser, but when I search for a question, it doesn’t look through the internet for results. Instead, it looks through a database I created with the names of my phone contacts and the knowledge areas I attribute to each of them (e.g., Dan: [woodworking, fishing, repair, cars, football, antique markets]). It compares the words in the question with the knowledge areas of each person, then displays the name of the person with the most relevant knowledge as the search result. For example, how do I make an egg casserole? returns Anya as the top match because, amongst other things, I said she has knowledge in baking, cooking, and farms. (For those interested in a more detailed technical explanation, see Appendix).

Figure 1. A screenshot of Ping’s interface with annotations on the various features.

For 5 days, I used Ping as I would use my usual search browser and tracked various metrics including time to receive a response, number of nodes to receive the response, usefulness of the response (1-5), emotion while sending the question (1 very hesitant, 3 neutral, 5 very willing), and emotion when receiving response (1 very negative, 3 neutral, 5 very positive). 

[Madu in reply, 26-02-26 13:10] They’re bitter oranges, usually for jam.

I think the ayuntamientos usually organize the community to pick them up and make them into preserves and jam.

Unlike modern search browsers and the brain, Ping employs a very naive algorithm to determine which of my contacts has the most relevant knowledge on the question. There are many flaws to its simplicity. For example, it breaks down the question into individual words, so phrases that have a compound meaning (such as ice cream, full moon, washing machine, real estate, hot dog) would be interpreted as individual words. If I ask Ping, “Where can I get an ice cream?”, it would look for people with knowledge in “ice” or “cream” and might return Evan who has knowledge in “skiing” and “snow” (because of ice) instead of Anya who has knowledge in “desserts”. While I could’ve spent more time refining the algorithm to be “smarter,” I decided that this lack of refinement was a more interesting highlight on how many factors our brain is automatically taking into consideration when we make decisions (language, context clues, situation, emotion) and the ones search engines seek to replicate (along with the many it can’t know or account for).

[Me, 28-01-26 08:29] Do you know if the store interior de te in Barcelona sells assam tea?

[Ping] Send text to Vaani.

Through this process, I anticipated I’d be most interested in the responses I received or even to be frustrated by the lack of immediacy in receiving the answers, but instead I was most fascinated and even discomforted by the process before sending the texts. In assigning knowledge areas to each of my phone contacts, I felt odd distilling my friends and family down to keywords. My mom is so much more than “California,” “biology,” “blood,” “real estate,” “motherhood,” “books,” etc. Even in the magnitude of the word “motherhood,” there feels to be an expansive gap between “keyword-her” and her. There were also long-term friends who I value deeply and who I turn to for various questions and advice but who I struggled to think of more than 4 knowledge areas for. Conversely, there were people I’ve just met who I could quickly list 10 or more knowledge areas for. Yet in all of it, I was inspired by the breadth of knowledge held by the people I know.

Figure 2. There was no relationship between time and emotional sentiment. But even when I got a response that didn’t fully answer my question, my emotions trended towards positive; maybe because I was happy to have the out-of-routine interaction with them.

Ping rarely gave me a result that was the usual person I would contact with the given question. It wasn’t because the algorithm was “incorrect”; instead, it revealed to me the ways I more heavily weight existing emotional connection over expertise when I’m reaching out to someone I know. Ping forced me to reach out to people I usually wouldn’t, and as a result sparked new connections and common interest points between me and them. The questions I asked often required the other person to seek out some additional information, and they almost always responded to me with additional links, insights, or even offers to find out more. Ping brought me out of my routines to remind myself that efficiency to an answer isn’t everything. I couldn’t tell you off the top of my head many, if any, of my Google searches from the week before my experiment, but I can certainly recall the conversations and interactions I had as a result of asking questions to friends, family, and strangers.

[Vaani in reply, 28-01-26 09:06] Hii, sorry I have no idea

I can try to look for it if you want some

[Me, 24-01-26 22:23] Are there any cool design events coming up in Barcelona?

[Ping] Send text to Martina.

[Martina in reply, 24-01-26 22:23] actually i don’t but i can ask my friendssss!

i want to too so I’ll let u know

[Martina forwarded link, 25-01-26 17:34] https://www.casamontjuic.com/agenda/

they have like little exhibitions

[Me, 25-01-2617:36] Oooo cool I’m going to look at the schedule! Thank you :-) :-) 

I’ll lyk if there’s one I’m going to

Appendix

Ping is built using Next.js for the frontend, Supabase for database management, and Hugging Face language model for natural language processing.

When the application loads, the “people” (my phone contacts and their knowledge areas) are fetched from Supabase.  The knowledge areas of each person in the database are converted into semantic vector embeddings (numerical representations of a word’s meaning) using the Hugging Face Sentence Transformers model and cached to improve performance.

When a user submits a search query through the UI, the application first processes the input by removing unnecessary words such as “and,” “of,” “for,” etc. The query is then assessed against the knowledge areas of each person to find which person has the most relevant knowledge. Two techniques are employed for this matching. The Hugging Face language model generates a semantic embedding of the query, which is compared against the semantic embeddings of each person’s knowledge areas using cosine similarity. Secondarily, Fuse.js performs text-based fuzzy matching to account for minor differences like typos. A weighted algorithm uses the scores from the cosine similarity and Fuse.js search to generate a final score for each person, 0-1 with 1 being a perfect match. Finally, the top result is returned to the user.