Azeem Azhar is the kind of guy who loves both old-fashioned pens and advanced AI.
It was a delight talking with him, not just because he’s a successful entrepreneur, author, and interviewer, but because we share quirky tech tastes.
Azeem and his team publish Exponential View — a Substack with 140,000+ subscribers — about how tech is shaping our future. In our live conversation, we talked about Azeem’s AI — and analog — workflow. The discussion also touched on 18 sites, apps, and gadgets summarized below.
📺 Watch the video for the full chat, or check the highlights and tool list below.
🔎 On AI Research Beyond Google and Wikipedia
Azeem consistently tweaks how he figures out where tech is heading. Check his “Boom or Bubble” dashboard on whether the AI market is overheated for an example of his analysis.
Key takeaway: Azeem’s research workflow has moved almost entirely away from traditional search:
* Why he quit Google and Wikipedia. [16:20] “I spend virtually no time on Google searching for things, nothing on Wikipedia at all, not a moment now.” Instead, he surfaces what matters to him with AI searches and custom tools his team has developed.
* Manus is his research assistant. Azeem and his team use Manus (a Chinese-founded AI startup recently acquired by Meta) the way you’d send a research assistant to “find color—go off and find the case studies, the anecdotes, the famous quotations.” The team runs queries overnight for 30-40 minutes each.
* Shortwave helps with investor intel. This AI tool is superb for searching within your email. Azeem has 40 startup investments. To support founders, he uses Shortwave to search past email to help explore questions like, “How well are they sticking to their milestones? Have they changed the goalposts? Where do they seem to have problems where I can be helpful?” Azeem uses Shortwave to search across 15 years of Gmail messages.
* Julius is a resource for data science. He uses this tool, which he’s invested in, as an AI number cruncher. [My post on Julius]
🤖 How Azeem Uses AI
Azeem set up custom instructions directing ChatGPT to aggressively challenge his assumptions: [6:41] “It can be quite exhausting… it’s like being constantly interrogated.” His follow-up? “I often have to copy the answer and put it into Claude and say, explain this to me like I’m a bright high schooler.” He considers Claude the best coding model and also uses it for text refinement.
Bottom line: Azeem uses the two AI models sequentially to force himself to think deeper.
Other AI Assistants
* Gemini Pro Azeem builds interactive apps with Gemini. “I might explain what I’m looking for, have a discussion, then ask it to build the interactive platform app. Then I can play with the parameters.”
* Perplexity Azeem’s go-to for “instant answers” when he doesn’t need deep research.
* DeepSeek Azeem’s default for “good enough” queries to cut costs. “In general, it’s really good enough. And if on the occasion I don’t think it is, I can fire it out to one of the other models.”
* Grok Azeem experiments with this occasionally as part of his testing.
✒️ “I Dip the Pen in this Bottle of Ink”
Azeem has a fountain pen without an internal ink cartridge. Why?
[9:30] “If I am writing and every 10 or 15 words I have to stop to dip the pen in ink, I’m slowing myself down... In a world where I can move really quickly, I will slow myself down. I’ll get very haptic in the experience and look at what I’m writing and force myself to cross out mistakes, and feel frustrated about mistakes, so therefore slow my thinking down even more.”
In a conversation about AI acceleration, Azeem deliberately builds in friction.
🎙️Voice and Writing Tools 🖊️
* Wispr Flow Azeem reads his handwritten notes aloud into Wispr Flow, editing in real-time as he speaks.
* Kolo Tino Fountain Pen Azeem likes the feel of pen on paper and its deliberate pace.
* Paper Republic Trifold Leather Journal Azeem’s folio holds three separate notebook inserts: “You can have one that’s just for your jotting of your to-do list and then others are for thinking time.”
😰 Azeem: “Are You a Bit Stressed, Jeremy?” 😖
In our recent conversation, Azeem teased me for repeatedly referencing resources for relaxation.
“Are you a bit stressed? Because you’ve talked to me about your squeeze ball. You’ve talked to me about Headspace for meditating. You’ve talked to me about your CMY cube to chill you out. I don’t want to go all shrink on you, but there’s a little hint of intensity there.”
Fair point. I do have a lot of calming tchotchkes on my desk. We both shared a bunch of tools we use, analog and digital, for coping with busy-ness and overwhelm. Below are three Azeem recommends:
Focus & Wellness
* Pzizz Azeem’s most-used app: “[The app’s] run time is probably 10 hours a week on average and has done for a decade.” He paid $50 for a lifetime subscription and uses it for naps, overnight flights, and jet lag recovery.
* Oura ring Azeem uses this for health tracking, as have I, since I first wrote about it in 2021.
* Lacrosse ball Azeem’s essential travel item: “A really hard lacrosse ball to put into your pressure points, like your glutes and your hip flexors, because you want to do that massage, especially once you’ve been on your second 10-hour flight in five days.”
🎧 High-End Audio: Headphones and DJ Mixing
Azeem tests headphones with “Sultans of Swing,” a song by Dire Straits, Bach’s Goldberg Variations, as well as dance tracks.
* Solitaire T Headphones Azeem travels with these for high-end sound with noise cancelling for long flights.
* Beats Fit Pro His everyday choice.
* DJ Studio Azeem uses this digital audio workstation to create DJ mixes while traveling.
🌟 Notable Quotes
* Azeem on the skyrocketing cost of cutting-edge AI: “All of this is expensive. I mean, we spend as a team, I would say probably between $350 and $500 per person per month.”
* My take: we’re nearing a new digital divide. Free alternatives like Jan AI help (see my post), but if you use the best contemporary AI models, your bills add up fast.
* Azeem on why no single AI company will “win” the market. "I haven't seen a strong reason to think that a network effect will emerge.”
* His rationale: “If you get the unique data for clown makeup, and I get the unique data for vineyards in Italy, if I'm not interested in clowns, I'm not going to use your LLM.”
* Azeem says OpenAI’s strategic dilemma, as reported by The Information, is whether to focus on power users or those who just want quick replies.
* [29:23] “They have 1,000 researchers who have built amazing deep research tools, and most people just want to know which movie should I watch tonight, or draw my dad as if he was Spider-Man for his 70th birthday card.”
* On accountability in the AI era: “We’ve essentially said to the team that we don’t have hallucinations at Exponential View because everything is owned by a person. If you copy and paste something out of an LLM that’s got a hallucination in it, that’s you saying it, that’s not the LLM saying it.”
* On the swift pace of change: “I would think that 70% to 80% of my workflows today, the things that I do, are different to how I did them three months ago.”
Thank you to Romaric Jannel, Abbey Algiers, Aysu Kececi, Ondrej Prostrednik, blaine wishart, and many others for tuning into the live video conversation with Azeem Azhar.
You’re Invited! Join me next Wednesday, January 21 at 10am EST for Just Start Writing: Tools For Busy Creators In 2026, a free workshop I’m leading for Medium.com.
This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit wondertools.substack.com/subscribe