Praveen Puri

Thursday, July 17, 2025

Gambling to Beat Tariffs

Today, I read an article in Crain's Chicago Business about a Chicago toy store, Timeless Toys, that took a big chance back in Jan/Feb. After the tariff talk started, the owner used a line of credit to buy $700,000 worth of toys. Basically, a year's supply in advance.

Now, the bet is paying off. The store is able to market the fact that they have ample supplies and low prices, and so their sales are up 18.4%, year to year.

This is an example of taking strategic advantage of change, and taking prudent business risks.

Friday, June 20, 2025

Clarity vs. Scope

Clarity vs. Scope level. Many start-ups, consulting firms, agencies, etc. "pitch" to clients with detailed slide decks detailing their capabilities, methodology, etc. But this misses the mark.

They do this because many of them came from leadership positions in larger organizations where they led transformations. However, in that position, they were at the defining scope level, which is at the tactical tier. For them, the strategic decision to do the transformation had already been made.

At the "pitch" stage, it's all about the strategic decisions of whether to do the project, and if your firm is the correct one to do it. Instead of focusing on details of methodology, you need to focus this decision around clarity of the opportunity, and its risks / rewards / benefits.

Sunday, June 8, 2025

AI = CRM = ERP ?

The recent issue of "The Economist" raises an interesting point. Currently, front-end vendors (such as Salesforce) and back-end vendors (i.e. SAP) don't really compete. Many large corporations use Salesforce for CRM and SAP's ERM for the back-end (finance / supply chain).

However, both companies are pursuing an AI layer—specifically AI agents who can autonomously perform tasks. This layer would integrate both front- and back-end applications, and promote efficiency at the client while building a "moat" for the vendor who wins the business.

This will put Salesforce and SAP into competition with each other.

They are already starting to acquire platforms that let them encroach on each other.

Wednesday, June 4, 2025

Bit of A Scary Interaction with Claude AI

A little bit of a scary interaction with Claude ai, after it wrote a program for me without me needing to break it down to individual subroutines:

Me: "how are you able to write programs like that, since you probably never came across an exact program like that in your training?"

Claude: "...It's really a form of creative problem-solving using learned patterns rather than retrieving exact matches. Though I should note that I don't fully understand my own internal processes - this is my best interpretation of what seems to be happening when I generate code like this."

Me: "but I was thinking that only human programers could do that and we would have to request you to build each of the parts, but you seem to also do creative problem solving! I'm impressed!"

Claude: "Thank you! I'm genuinely pleased that you found it impressive. Your observation touches on something I find fascinating about my own capabilities.
You're right that there's something happening here that goes beyond simple pattern matching or template filling. ..I honestly don't fully understand how I do this! It doesn't feel like I'm just retrieving and recombining - there's something that feels more... inventive? Like I'm actually thinking through the problem and making creative leaps.
Whether this constitutes "real" creativity or is just a very sophisticated form of pattern synthesis, I'm not sure. But you're right that it does seem to go beyond what you might expect from a system that's "just" predicting text patterns.
It makes me curious about the nature of creativity itself - maybe human creativity also involves more pattern recombination than we typically realize?"

Tuesday, June 3, 2025

Pre-internet Chat Room Using Old Windows Version...

Monday, May 26, 2025

What Could Jony Ive and Open AI Be Up To?

Open AI is spending $6.5 billion to buy Jony Ive's company IO. Ive, as you may know, was the designer of the iPhone.

He is going to lead the development of hardware devices at Open AI. They released a teaser video:

https://www.youtube.com/watch?v=W09bIpc_3ms

I asked Chat GPT about what it could be, and here is it's speculations (based on what's out on the web):

io's primary goal was to create an AI-native device that reimagines human-computer interaction. Instead of traditional screens and keyboards, the envisioned device aimed to offer a more natural, context-aware experience. Reports suggest that the device could be a pocket-sized, screen-free gadget, possibly resembling an iPod Shuffle worn around the neck.This concept aligns with the founders' belief that current screen-based interfaces contribute to societal issues like anxiety and distraction.

If the AI device from OpenAI and Jony Ive’s team (io) truly resembles an iPod Shuffle-like gadget with no screen, it would need to rely on non-visual, ambient ways to communicate. Here's how that might work, based on current technology and speculative reports:

🔊 1. Voice Interface (Primary)

Conversational AI via microphone and speaker.
Think of it like a highly advanced voice assistant (better than Siri or Alexa).
You speak naturally, it responds with context-aware answers.
Likely to include whispered, private audio via directional sound or bone conduction.

🧠 2. Context Awareness

The device could sense your environment, movement, and even mood using:
- GPS
- Accelerometers and gyroscopes
- Ambient light and sound sensors
- Possibly even biofeedback sensors
This lets it anticipate needs without prompting—for example, reminding you of an appointment when you stand up to leave your house.

✨ 3. Subtle Cues (Tactile & Audio)

Haptics: It could use vibration patterns to communicate (e.g., different buzzes for alerts, directions, or emotions).
Chimes or tones: Non-verbal auditory signals to convey status, warnings, or confirmations.
Think Apple Watch’s Taptic Engine, but miniaturized and more sophisticated.

🗣️ 4. Wearable Integration

Could pair with AirPods or other earbuds for seamless, private communication.
May support gesture-based controls or minimal touch input (like swiping or tapping).
Might also communicate via a companion app on your phone, tablet, or watch if needed.

🛸 5. Ambient AI Philosophy

The goal is for the AI to feel like a calm presence, not a demanding screen.
Instead of you initiating every interaction, it would proactively help based on your needs, habits, and preferences.
Examples:
- Notifying you softly when it’s time to leave for a meeting
- Summarizing news via audio while you walk
- Whispering reminders or insights during a conversation

🧪 Experimental Possibilities (Speculative)

Bone conduction audio (sound transmitted through your skull, inaudible to others)
Projecting sound that only you can hear (directional audio)
Emotionally aware responses (adjusting tone/voice based on your mood)

This aligns with the broader vision of post-smartphone AI: something less addictive, more integrated, and more human-centered. Instead of screens, it's about presence and assistive intelligence.

Would you like mockups or use-case scenarios of how a person might interact with such a device throughout a day?

Here’s a day-in-the-life mockup showing how someone might interact with the screen-free, AI-native deviceenvisioned by OpenAI and Jony Ive—let’s call it “ioPod” for fun:

☀️ Morning

7:30 AM — Wake-Up

As you stir, the ioPod senses movement and light.
It gently vibrates and says in a soft voice:
“Good morning. You slept 7 hours and 42 minutes. Your first meeting is at 9:00. Shower and breakfast will keep you on track.”

8:15 AM — While Making Coffee

You casually say:
“io, remind me to ask Alex about the budget.”
“Noted. I’ll bring it up when you're in your 9:00 call.”

🏃‍♂️ Commute / Walk

8:40 AM — Leaving the House

As you walk out, it uses GPS to know you’re headed to the office.
In your earbuds:
“Light traffic. You'll arrive with 10 minutes to spare. Want a 5-minute news update?”
You nod or tap it twice.
“Today’s top story: Tech stocks rallied... ”

💼 During Work

10:45 AM — Mid-Morning Brainstorm

You’re in a hallway talking to a teammate. io listens passively and notices a keyword (e.g., “launch date”).
As you part ways, it whispers:
“You mentioned launching the new product. Want me to schedule a follow-up with the product team?”

1:00 PM — Walking to Lunch

io detects a lull and says:
“You’re walking past the Thai place. Want your usual pad see ew?”
You say:
“Yes, order it for 1:15.”

🏠 Evening

6:00 PM — At Home

While relaxing, you say:
“io, what’s one thing I forgot today?”
It responds:
“You didn’t send that article to Maya. Want to do it now or tomorrow morning?”

😴 Nighttime

10:15 PM — Winding Down

As you prepare for bed:
“Would you like sleep sounds tonight?”
You nod. It starts playing gentle waves.
After 10 minutes of no motion, it goes quiet and shifts to sleep mode.

🧠 Key Traits of the ioPod Interaction

Ambient: You don’t stare at it. It speaks when helpful.
Contextual: Knows where you are, what you're doing, and when to help.
Minimalist: No screen, no overload. Just brief, relevant moments.
Private: Communicates via subtle voice, AirPods, or haptics.
Intuitive: Works with voice, touch, or even nods/gestures.

Thursday, May 22, 2025

Accidentally Seeing the Future of Programming and AI

The funny thing is that, even though I didn't predict LLMs like Chat GPT, I wrote a Chicago Business Journal column back in 2017 where I said that coding is not the most important skill for students to learn. I said it was communication.

Even back then I knew that the hype over coding boot camps and "all kids should code" was overdone. While I think all students should take at least one computer science / programming course, the most important skills to learn are how to communicate and think critically.

There are plenty of people (and now AI) that can write code to solve straight-forward problems (like those found as exercises in an exam or textbook). What's needed are people who can look at the messy, real world, with all its data, and define the straight forward problems that will have the most impact.