Review of Q4

At TYPE III AUDIO, we narrate important writing and help people find the best thing to listen to. We're Spotify, for people who want to solve the world’s most pressing problems. We started in October 2022. This is a review of our first three months.

Strategy updates

  • We started with the plan to create 500 hours of human narration within 1 year.
  • We looked into AI narration. We realised that it is quite good and improving rapidly. So we decided to also release at least 1000 hours of AI narration during year 1.
  • From the start, we knew that marketing would be important. There is no point making narrations if nobody listens.
  • We started thinking more about marketing. Diverse clients will require diverse approaches, but there will be common patterns. We will develop “playbooks” based on initial projects for some pilot clients. One worry: marketing the narrations of a particular client may lead people to listen to those narrations when, in fact, there are much better things for them, personally, to listen to.
  • We continued thinking about marketing. We realised that we should try building our own platform: TYPE III LIBRARY. The aim of the platform would be: help people find the next best thing to listen to. If we build a great discovery platform, we increase the reach of our narrations, while also improving the value people get from their average listening session.

Our focus

Our focus in Q4 was “learning and investment”. Key parts of that:

  • Talking to listeners and clients
  • Setting up our human narration and audio editing services
  • Building our AI narration pipeline
  • Building the TYPE III LIBRARY
  • Hiring
  • Setting up the company (e.g. critical operations; branding).

Roughly 4/5 of our staff time was spent on “learning and investment” activities like these.

The remaining 1/5 of staff time was spent on delivering services to our clients.

Hiring

We started in October with:

  • Peter Hartree (CEO)
  • Jason Cotrebil (Audio engineer)
  • Katy Moore (Head of QA)
  • Perrin Walker (Narrator)

This team had previously worked together on Radio Bostrom, which can, in retrospect, be seen as a pilot for TYPE III AUDIO.

We trialled and hired five people during Q4:

  • Anna Dilja (Executive Assistant)
  • Tom Norton (Designer & developer)
  • Uta Reichardt (Narrator)
  • Marcus Sommers (Quality Assurance)
  • Igor Stajic (Developer)

The average team size during Q4 was roughly 2.5 FTE.

At the end of December, the team size was ~4.5 FTE.

All staff are currently employed as contractors, working remotely. Most are gradually increasing their time commitment to TYPE III AUDIO (e.g. Peter Hartree and Perrin Walker are now full-time).

We also hired And-Now, a creative agency, to work on branding and web design.

Finally, we confirmed three “informal advisors” to the project: Fin Moorhouse, Joseph Walker, and Valgeir Valdimarsson. We will seek a couple more during Q1.

We will continue hiring in Q1. We have three further work trials scheduled for January.

AI narration pipeline

The key elements of our AI narration pipeline were built. The system can automatically extract text from any URL, apply automatic transforms, and then put the narration script into a Google Doc for manual review.

In January, we will test the system heavily, building up our database of automatic transforms and testing the best text-to-speech services. We expect to begin testing with client projects by the end of the month.

TYPE III LIBRARY

We began the design and development of the TYPE III LIBRARY. The first rooms will open to select visitors in January. We will open to the public sometime in Q1.

Client projects

As mentioned above, roughly 1/5 of our staff time was directly spent on delivering work for clients.

We worked with 5 clients, namely:

  • 80,000 Hours
  • Effective Altruism Forum
  • LessWrong
  • Joe Carlsmith
  • Joseph Walker (Jolly Swagman Podcast)

Human narration team

Some examples of the work we completed:

In numbers:

  • Narration episodes released: 52
  • Total narration time: 31.5 hours*
  • Narration downloads: >40,000
  • Estimated listening time: >10,000 hours** >5,000 hours***

*Industry standard is that it takes 3-6 hours of narration and editing work to create 1 hour of finished audio.

**This figure assumes that episodes are 0.5 hours on average (about right), and that 50% of episodes downloaded are listened to in full (pretty uncertain).

*** Estimate updated 2023-09-14.

“You record, we everything” team

We set up the EA Forum Podcast on behalf of the EA Forum Team. From a very high-level brief, we scoped the project, created the feeds and cover art, got them listed on podcast directories, co-ordinated with external stakeholders. We are now running the podcast on their behalf.

We expect to offer a “podcast creation and management” service to other clients in Q1. Katy Moore has expanded her role to become “Podcast Manager” in addition to “Head of QA”.

Jason Cotrebil took on our first two clients for the “You record, we edit” service. We edited ~15 episodes during Q1.

We took over Radio Bostrom, the pilot project which led to TYPE III AUDIO. With only a modest launch, the project has seen ~13K episode downloads in the first 4 months. Key stakeholders are delighted, and we raised $25K to enable a big marketing push for the project in early 2023.

Client feedback and testimonials

Client feedback has been very positive so far, though there are many aspects of our services we are keen to improve. Several of our early clients have already agreed to share public testimonials.

TYPE III AUDIO has produced high-quality narrations of our articles that have increased their reach for comparatively little effort. Working with them has been incredibly smooth  for us, and they keep coming up with attractive additions to their services we’re interested in too!

– Arden Koehler, 80,000 Hours


TYPE III AUDIO's efficiency and professionalism make them wonderful partners. By partnering with them, we’ve been able to test audio narration on the EA Forum quickly and develop a strategy to boost engagement further.

– Sharang Phadke, Effective Altruism Forum

TYPE III AUDIO created a full, 3+ hour, high-quality narration of a report I wrote about power-seeking AI, and they've been helping improve the quality of the narrations of other things I've written as well. In all of this, they've been enthusiastic, detail-oriented, communicative, serious about the quality of their product, and keen to be as helpful as possible. I've been very satisfied with the narrations, and it's been great working with them more broadly.

— Joseph Carlsmith, Open Philanthropy

Branding & web design

And-Now began work on our brand and visual identity. They developed a promising brand concept, and delivered our first company logo. In January, they'll finish our "year 1" identity, and help us with UI and UX design.

TYPE III AUDIO brand concepts
Illustrative UI mockup

Progress against our stated plan

In October, we summarised our Q4 focus as follows:

  • Get started
  • Test & decide what services we should offer in year 1 (and pricing)
  • Complete some initial projects and delight our first clients
  • Setup and refine narration workflows, sales pipeline
  • Hiring

We did all of the above, to a standard I am happy with. We did quite a lot else, as well.

Challenges we faced

In November, I was quite distracted by other commitments, especially my advisory role at 80,000 Hours. This was due to the fallout from the FTX disaster. I clocked up more than 50 hours of work related to FTX, which was more than 5x my usual monthly commitment. I’m now back to the usual 5-10 hours / month and do not expect similar issues in 2023. I was pleased to see that the TYPE III AUDIO team maintained a high rate of progress during this period. The main cost was that my error rate was higher than usual, due to extreme speedrunning. This led to one medium-size management mistake—neglecting a relatively new hire—which cost us £1-5K.

We also made some relatively minor mistakes in our work on hiring and client projects. I’m encouraging the team to speedrun lots of things, and I’m happy with our rate of progress, and our error rate. I’d actually be happy to see our error rate increase (in relatively low-stakes areas), if it means we can move faster.

Plan for 2023

Outline:

2023 Q1: We’ll focus on learning our way to product-market fit.

2023 Q2: We’ll get to ramen profitable, or seek external funding.

2023 Q3: We’ll end our first year with one or more of:

(a) Monthly profits in August and September.
(b) Net annual profit of +/- 20%.
(c) At least 12 months’ runway.

Our overall vision for the project continues to evolve. The latest version is here.

Plan for 2023 Q1

We will:

  • Develop our AI narration pipeline and release >100 hours of AI narration.
  • Launch TYPE III LIBRARY, a website that will help people find the next best thing to listen to.
  • Release >10* hours of human narration.
  • Release >10* episodes via our “You record, we everything” service.
  • Grow and invest in the team. As part of that, we’ll run our first team retreat in early March.

*These targets are very conservative. We are setting them low because we don’t want to overcommit to client projects during Q1. If we wanted to maximise these outputs, we could probably deliver 10x these numbers.

Closing thoughts

Our first 3 months have gone well. Our vision for the project has evolved a lot, and I have decided to make it my main focus for 2023.

I am especially delighted that we narrated some 50 articles, all discussing topics related to the world’s most pressing problems. And people were listening—our narrations received at least 40,000 downloads during the period.

I’d like to give a special shout out to Perrin Walker, Katy Moore and Marcus Sommers, who worked with great energy, focus and commitment to complete 52 narrations amidst the usual “startup just getting started” chaos.

Many of the investments we made during this period will start yielding fruit during the next 3 months. I’m excited to get into it.