04 May 2023

ChatGPT Releases Code Interpreter: Your Personal Data Scientist (#115)

2Frameworks

11Insights

Frameworks in this episode

Strategy5 steps

AI Community Defensibility Test

Separate replaceable Q&A marketplaces from communities built on belonging

Read →

Innovation5 steps

Five-Awareness AI Agent Scorecard

Evaluate AI agents across compute, data, coordination, safety, and user value

Read →

Insights & moments

The myth-busts, hot takes, explainers, and tools worth keeping.

Hot Take· 3

Hot Take04:00

Natural-Language Analytics Could Replace Fixed Dashboards

Conversational analysis changes reporting from one standardized view for many users into a personalized interface for each user. The hosts expect people to request exactly the chart or insight they need rather than navigate rigid analytics dashboards.

Reporting shifts from one-to-many to personalized output
Users can generate charts without specialist skills
Natural language reduces dashboard complexity
Presentations may be generated directly from analyses

“reporting user experiences, are largely in a lot of cases, gonna go away, 'cause people are just gonna chat query to get the insight that…”
Kipp Bodnar · 04:00

“Now I can actually just build the charts that I want through natural language.”
Kieran Flanagan · 04:30

#analytics#dashboards#natural language#reporting

Hot Take11:30

Human-Service Businesses Face the Greatest AI Pressure

Businesses whose core product is expert labor and billable human time may face particularly rapid disruption. AI can offer overlapping services faster and more cheaply, putting pressure on both talent marketplaces and service-heavy public companies.

Human expertise marketplaces are exposed
Billable hours become less defensible
Faster and cheaper AI alternatives increase substitution risk
Knowledge-intensive public companies may warrant scrutiny

“the more you are dependent on human skill and hours to be successful as a company, the more disrupted you are going to be”
Kipp Bodnar · 12:00

“the more you are offering services to other humans that AI can offer faster, better, cheaper, man, that is a big part of the disruption…”
Kipp Bodnar · 12:00

#services#labor#automation#business strategy

Hot Take20:00

AI Writing Tools Help Research but Rarely Produce Shippable Work

Kieran argues that current AI writing products are useful for research, outlining, and fragments of a first draft but do not consistently produce excellent finished content. Code Interpreter appears more transformative because its present-day analytical output is closer to business-critical utility.

AI can accelerate research and outlining
Generated drafts often require substantial human work
Good demonstrations do not equal shippable output
Some AI use cases are currently more consequential than writing

“AI Writing tools they're good, but they're not life changing.”
Kipp Bodnar · 20:30

“But is it gonna write the best piece of content, is it gonna write a great sales page, is it gonna be the best copywriter…”
Kieran Flanagan · 20:30

#ai writing#content#hype#productivity

Explainer· 1

Explainer01:00

Code Interpreter Puts a Data Scientist in Every Chat

Code Interpreter lets users upload data and interrogate it through natural language without first mastering analysis tools or Python. It can inspect datasets, answer follow-up questions, generate charts, and adapt an analysis to changing requirements.

Upload spreadsheet or CSV data directly
Query datasets in natural language
Generate analyses and visualizations
Revise results through conversational follow-up

“I think Code Interpreter, we can just summarize it as now you have a data scientist with you at all times.”
Kieran Flanagan · 01:00

“all you need now is data, right? Like any kind of data extract and you can upload files, it can download files.”
Kieran Flanagan · 02:00

#code interpreter#data science#chatgpt#analytics

Story· 3

Story05:30

Chegg's Collapse Shows AI Disruption Is Already Here

Chegg's weak earnings and sharp market-cap decline are presented as evidence that ChatGPT is already changing commercial behavior. The example challenges the assumption that meaningful AI disruption remains months or years away.

Chegg provided tutoring and educational support
ChatGPT substituted for part of its service
The company cited ChatGPT when reporting weak earnings
AI disruption is affecting businesses now

“their market cap got cut in half, the stock dropped in half when they reported bad earnings and cited ChatGPT as the disruptor and cause…”
Kipp Bodnar · 06:00

“this stuff is like right now, and changing the business landscape and dynamic right, right now.”
Kipp Bodnar · 06:00

#chegg#education#disruption#business

Story15:30

Walmart's Negotiation Bot Cut Expenses by Three Percent

Walmart tested Pactum AI as an automated negotiator with a subset of vendors and reportedly achieved a three-percent expense saving. The result illustrates why large enterprises may adopt AI quickly despite concerns about errors and organizational disruption.

Pactum AI negotiated with vendors
Walmart began with a limited test
The test reportedly reduced expenses by three percent
Successful economics encouraged a wider rollout

“it found that it saved that they're on three percent on expenses, and so now it's gonna roll it out to most of its vendors.”
Kieran Flanagan · 16:30

“A three percent expense saving is massive.”
Kipp Bodnar · 17:00

#walmart#negotiation#enterprise ai#cost savings

Story22:00

Midjourney 5.1 Demonstrates the Speed of AI Improvement

A side-by-side comparison using the same prompt showed a meaningful visual improvement from Midjourney 5 to 5.1. The hosts use the example to argue that current shortcomings may disappear in months or weeks rather than years.

The same prompt produced better output in version 5.1
A small team was shipping major improvements rapidly
Model limitations can become outdated quickly
Adoption decisions should account for accelerating capability

“the feedback, like the cycle of innovation is so fast on this.”
Kipp Bodnar · 22:30

“it's not years”
Kipp Bodnar · 22:30

#midjourney#image generation#innovation speed#models

Tool· 2

Tool15:00

Combine Web Scraping with Code Interpreter for On-Demand Research

Pairing scraping with Code Interpreter creates a powerful research workflow: collect structured information from a website, then question and analyze the resulting dataset conversationally. This expands Code Interpreter from user-owned files into broader web-derived analysis.

Scrape data from relevant websites
Collect the results into a structured dataset
Analyze the extract with Code Interpreter
Ask iterative questions without manually coding each analysis

“You can scrape any website, collect the data and then run it through Code Interpreter and ask the questions”
Kieran Flanagan · 15:00

#scraping#code interpreter#research#data analysis

Tool24:00

BabyAGI Turns Broad Goals into Recursive Task Loops

BabyAGI can translate a broad objective into tasks, execute them, use the findings to generate additional tasks, and repeat for a chosen number of iterations. Its results remain uneven, but the improving interface and loop behavior show how quickly autonomous-agent experiments are progressing.

Start with a broad goal
Let the system generate a task list
Execute tasks and create follow-up work recursively
Limit iterations and API spending
Treat current output as research rather than finished work

“BabyAGI is basically you can build recursive loops”
Kieran Flanagan · 24:00

“if you're using OpenAI keys, do set soft limits, so you can say I wanna spend max a $100 this month, do not get caught…”
Kieran Flanagan · 25:00

#babyagi#agents#automation#recursive loops

Takeaway· 2

Takeaway02:30

AI Turns Existing Digital Assets into Interactive Knowledge

The hosts contrast AI with technologies that require users to create new assets or behaviors. AI can work retroactively on existing documents, datasets, and workflows, making established resources easier to query and more valuable.

Existing documents become conversational
Legacy digital assets gain new utility
AI adoption can build on familiar workflows
Accessibility increases the value of stored knowledge

“What AI is doing is, is making all of your digital assets interactive.”
Kieran Flanagan · 02:30

“the brilliance of AI is it's retroactive, right? It's taking all the stuff that everybody already uses and making it 10 times better.”
Kipp Bodnar · 02:30

#digital assets#documents#knowledge#ai

Takeaway08:00

AI Democratizes Skills Once Limited to Specialists

The hosts frame AI as a broad democratization of capabilities, extending beyond the publishing and connection gains of Web 2.0. People can increasingly analyze data, visualize findings, code, design, and publish without the specialist resources previously required.

Specialist capabilities diffuse across the population
Old resource constraints become less relevant
Early adopters gain an advantage in applying new capabilities
The limiting skill becomes choosing and using the right capability

“Web 3.0 where the AI era democratizes everything.”
Kipp Bodnar · 08:00

“it's basically gonna be who can use the Ironman suit the best.”
Kieran Flanagan · 10:00

#democratization#skills#early adopters#innovation