08 August 2023

These 9 A.I. Tools Are Better Than ChatGPT… (#145)

2Frameworks

11Insights

Frameworks in this episode

Finance6 steps

AI Investment Durability Scorecard

Evaluate AI companies by distribution, economics, defensibility, and platform exposure

Read →

Marketing5 steps

Signal-and-Pivot Content Loop

Test topics, follow audience signals, and concentrate effort on proven demand

Read →

Insights & moments

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

Hot Take· 1

Hot Take16:30

AI Could Make One-to-Ten-Person Businesses Far More Profitable

Kipp Bodnar argues that AI and automation will enable more highly profitable companies with only one to ten employees. He suggests a dedicated venture strategy could target the software and automation providers serving this historically difficult segment.

AI may increase the capability of very small teams
Higher profitability could reduce traditional small-business failure risk
A focused fund could back tools serving one-to-ten-person companies
The opportunity lies partly in enabling infrastructure rather than only the businesses themselves

“Because of AI and automation, we are going to have more and more profitable, more successful one to 10 employee businesses”
Kipp Bodnar · 16:30

“the profitability of that category is going to improve.”
Kipp Bodnar · 17:00

#small business#automation#venture capital#entrepreneurship

Explainer· 4

Explainer09:00

Six Categories Organizing the Early AI Application Market

Kieran Flanagan maps the early AI application landscape into chat models, creative tools, B2B applications, data analysis, knowledge work, and productivity. The taxonomy provides a snapshot of where product creation and investor attention were clustering.

Chat products include ChatGPT, Bard, Claude, LLaMA, Poe, and Perplexity
Creative products span writing, image, and video generation
B2B products target sales, support, success, and go-to-market operations
Data products add conversational interfaces to spreadsheets and company data
Knowledge tools help employees retrieve information dispersed across company systems
Productivity tools focus on accelerating individual work

“one of the things I've been doing on AI is trying to like, map out where I see a lot of the momentum right now”
Kieran Flanagan · 09:00

“The fourth is data analysis, right?”
Kieran Flanagan · 10:30

#ai market#software categories#b2b#productivity#creative ai

Explainer17:30

Why Personal AI Models on Phones Could Become a Major Category

The hosts highlight models operating on edge devices as a distinct AI opportunity. Products such as Rewind could place a private assistant over a person’s own activity and data, creating a category with relatively few but potentially large winners.

Edge models can operate directly on personal devices
Personal data can make assistants more context-aware
Local operation may offer stronger privacy and security
The category may consolidate around a small number of major providers

“having personalized models in like, on your phone”
Kipp Bodnar · 17:30

“the handful of companies that win are going to win pretty big.”
Kipp Bodnar · 18:00

#edge ai#personal ai#privacy#mobile

Explainer28:00

Latency Is the Missing Piece for Convincing AI Sales Avatars

The group argues that AI avatars remain unconvincing partly because of the delay between a user speaking and the system replying. Wolfe describes a game-conversation system that processes speech while it is still being delivered, suggesting that streaming inference, realistic voice synthesis, and avatars could soon converge.

Long response gaps make AI conversations feel artificial
Convey reportedly processes early words while the speaker continues talking
Streaming processing can reduce the response gap to milliseconds
ElevenLabs-style voices could reduce robotic delivery
Avatar tools could add a human-looking visual layer
The components existed but had not yet been unified into one sales product

“the biggest problem with a lot of those tools is honestly, just the latency between when you ask it the question,”
Matt Wolfe · 28:00

“It's just the technologies all exist, I haven't just seen one tool pull 'em all together into a single”
Matt Wolfe · 30:00

#ai avatars#latency#conversational ai#sales#gaming

Explainer31:30

Company-Specific Models Could Handle Nearly All Support Requests

Wolfe predicts that organizations will train AI systems on documentation and historical support conversations. Those models could answer in the company’s established style and resolve the overwhelming majority of requests without a human representative.

Company data can customize general-purpose language models
Documentation supplies factual product knowledge
Past conversations demonstrate how issues were handled
Customer support is positioned as an early high-value use case
Human representatives would focus on the small residual set of difficult cases

“most large companies will have a plug and play version of a AI, like an LLM model just for them, right?”
Kieran Flanagan · 31:30

“these AI bots are gonna be able to handle 98.9% of the conversations that come through”
Matt Wolfe · 33:00

#customer support#custom models#enterprise ai#knowledge base

Story· 4

Story01:00

The Decade of Experiments Behind Matt Wolfe’s AI Breakout

Matt Wolfe traces his path from digital marketing and WordPress education to affiliate marketing, podcasting, and eventually AI coverage. His sudden growth in 2022 rested on more than a decade of accumulated content-production and marketing skills.

Wolfe began content marketing in 2009
Affiliate promotions generated much of his long-term income
Early projects included courses, blogging, podcasting, and WordPress tutorials
His eventual AI success built on years of uneven experimentation

“I've been doing digital marketing sort of stuff since 2009.”
Matt Wolfe · 01:00

“most of my income over the last decade has actually come from affiliate revenue from promoting other products in my content”
Matt Wolfe · 01:30

#creator economy#career#affiliate marketing#youtube

Story04:30

Future Tools Began as Matt Wolfe’s Personal AI Spreadsheet

Wolfe created Future Tools because a spreadsheet containing around 100 AI products had become difficult to navigate. Sharing the database publicly turned an internal organizational aid into a website attracting more than a million monthly views.

The project began as a solution to personal information overload
A spreadsheet became unwieldy at roughly 100 rows
A Twitter share helped the website spread
The site grew to over one million monthly views

“I literally built Future Tools as like a database for myself to kind of wrap my head around what tools I've come across and what…”
Matt Wolfe · 04:30

“now, it gets over a million views per month”
Matt Wolfe · 05:00

#future tools#ai tools#product origin#audience growth

Story24:00

The Voice Clone So Accurate Its Owner Couldn’t Recognize It

Wolfe recalls training ElevenLabs on his voice and initially believing the output did not sound like him. Viewers said they could not tell when the synthetic speech began, demonstrating both the technology’s quality and its potential for impersonation scams.

ElevenLabs could reproduce a voice from training audio
Wolfe experienced cognitive dissonance hearing his own clone
Viewers found the transition from real to synthetic speech difficult to detect
The same capability could support fraud and family-emergency scams

“I didn't realize you stopped talking.”
YouTube commenter quoted by Matt Wolfe · 24:30

“this is mind-blowing, this is groundbreaking, but also, this scares the hell out of me.”
Matt Wolfe · 25:00

#voice cloning#elevenlabs#deepfakes#fraud

Story34:30

Character.AI’s Appeal Became Clear Through Preteen Users

The hosts struggle to understand Character.AI’s valuation until Wolfe describes its popularity with his children and their peers. For middle-school-aged users, conversational characters function like a new generation of digital companions, though the adults worry about effects on real relationships and social skills.

Wolfe’s preteen children regularly used Character.AI
The product appeared popular among their similarly aged peers
Digital characters resemble an interactive evolution of Tamagotchi
Adult observers may underestimate demand because they are outside the core demographic
Synthetic companionship raises concerns about social and interpersonal development

“among like, their age kids, like middle school aged kids, they love this sort of like, fun, digital”
Matt Wolfe · 35:00

“It's like the new Tamagotchi kind of thing, right?”
Matt Wolfe · 35:30

#character ai#digital companions#children#social skills

Tool· 2

Tool23:30

Claude 2 Made Entire-Book Analysis Practical

Matt Wolfe describes Claude 2’s 100,000-token context window as a major practical breakthrough. The capacity made it possible to provide very long documents and ask questions, summarize material, extract bullets, or assist with rewriting.

Claude 2 supported a 100,000-token context window
Users could analyze books and very long articles
Wolfe preferred it for summarization and bullet extraction
Large context changed what could be handled in one interaction

“you can essentially upload an entire book, and ask questions of that book, right?”
Matt Wolfe · 23:30

“I've actually started using Claude 2 almost more than ChatGPT”
Matt Wolfe · 23:30

#claude#anthropic#long context#summarization

Tool26:00

Matt Wolfe’s AI Toolkit for Images and Video Effects

Wolfe names Midjourney, Leonardo.ai, RunwayML, and Kaiber among his favored visual-creation tools. Leonardo’s trainable styles and the video tools’ unusual effects are particularly useful for producing visuals that make viewers curious about the creative process.

Midjourney’s model updates repeatedly improved image quality
A web interface was being tested as an alternative to Discord
Leonardo.ai could train on collections of preferred images
RunwayML and Kaiber supported distinctive AI video effects

“every time Midjourney releases a new update, that just blows my mind.”
Matt Wolfe · 26:00

“You can upload a whole bunch of images that you like, and it will train on those styles”
Matt Wolfe · 26:30

#midjourney#leonardo ai#runway#kaiber#video