5 things I’ve used ChatGPT for in the last week

Issue 99 l Eka’s Weekly Roundup (26 November 2024)

Nov 26, 2024

We’re bringing you a slightly different newsletter today, which is a reflection on my (increasingly daily) use of ChatGPT. While I still don’t consider myself a super-user, and estimate >70% of my queries on ChatGPT to be work related, I’ve definitely noticed a shift towards non-work queries.

Given our consumer lens at Eka, I thought it was worth sharing how I’m increasingly leveraging GPT for everyday tasks. I remain very excited to see more applications built on top of this because, as we’ll see, there is still a long way to go for many use cases.

P.S. I’d highly recommend Meera Clark’s Redpoint Radar piece from her Conscious Consumer blog. This inspired a lot of writing.

GPT in our household 🗞️

Our household is now truly GPT converted. I counted up my daily frequency of GPT search frequency and, in the last 7 days, have averaged about 3-5x unique queries (per day).

Here are a few glimpses into (mostly) non-work related queries.

1: Asking ChatGPT if I can install a climbing board in a doorway

I recently ordered a Beastmaker and now need to go through the (~~pain~~) process of drilling this in our doorframe.

The problem is that:

I’m not a very good handyman,
My partner has already extensively helped around the house,
The closest climbing gym is 30m+ cycle away and 0C temperatures are a very strong incentive to stay home and train in the warm.

So this felt like mountain I actually could & should tackle.

I started uploading random images of doorframes in our (small) flat and asking it with overly simple prompts whether I could stick on the Beastmaker… see below for a few examples.

I then started uploading photos given the recent multimodality updates. My prompts are pretty basic and I find myself texting GPT a lot rather than putting in long thoughtful prompts (work examples are different, as you’ll see below).

I’ve cut the image above - note that it does give you a long thoughtful answer on this. To be honest I still think there is a way to go on understanding depth and lighting within the images as both frames are virtually identical. GPT feels like it has a bias to agreeableness rather than purely independent & critical viewpoints.

I then went onto Amazon to see if I could get a stud finder, and then realised my climbing friends probably had some themselves. I’m collecting it this weekend so I’ll let you know how I get on with this…

2: GPT (un?)intentionally roasting our living room when asked about how to make it prettier

I then asked GPT for some home decor tips. We’ve been trying to spruce up our pretty disorganised living room and were already shopping around on Ikea & Amazon for more storage containers. But hey, why not ask GPT for this too?

Once again, uploaded an image of the bikes (we’re also recycling an old suitcase which you can see below.. but GPT took that to heart, as you can see in its response).

I laughed a bit when it told me to declutter and organise… my partner agrees with GPT that our place is too cluttered. Win for him. Anyways, we now have a storage basket to keep our yoga matts. Not sure if GPT really pushed us to do this but it was a fun experience!

I think it can get a lot better on specific search. For example, we spent ages on Amazon & Ikea setting out the depth and width parameters (i.e. base diameter has to be 40cm but height can be up to 50cm because we have a plug socket in that little corner). I’m excited for GPT to better integrate with product catalogues so that your ‘nichest’ requests can be found without scrolling retailers for (almost) an hour.

3: Helping us understand the pros and con of getting an electric food composter

Just to say something upfront, we definitely didn’t end up purchasing one (cost, space, noise!). And it would add to the clutter.

For context, we wanted to get a new food waste bin and were looking at this… Kind of cool right?

We then thought about buying it and before long were asking GPT all kinds of things about electric foot composting. It was quite high level and wasn’t very salesy (I don’t think it could have sold us into buying it given our constraints) but still, it was a helpful sounding board!

We ended up going for a council food waste bin instead. A lot simpler.

4: Explaining FDA de novo regulatory processes

This is one example of a non-home related query. For more work related & technical searches, I typically start by asking a very simple question (below).

Unsurprisingly, I get a very simple answer. Then I try to explain my context and reasoning a bit more in the prompt. This was quite a back & forth conversation so I’ve left out some pieces below.

To be frank, I’d say 70%+ of my queries at the moment are work related rather than home related. There’s so much getting up to speed that we need to do that GPT has felt like a serious unlock here.

5: Very early use of voice in having it teach me German

Substack writing isn’t the best medium to talk about voice interactions, so here are some transcripts of how I’ve started speaking (very basic) German to GPT. I opted for the Juniper voice as it was my favourite when starting out.

Here’s a transcript of some ~~conversations~~ sentences exchanged with Juniper, in German. My partner tells me that her accent isn’t 100% there yet (he’s German) but to me I can’t really tell the difference.

This is where I felt the most need to prompt. It started out speaking way too fast and wasn’t properly explaining itself in English.

I would start asking it to define words and then it caught on that it should probably revert back into English if I didn’t understand specific words.

This is an example where a dedicated consumer application still makes to me. As you’ll know, I’m a super Duolingo fan and humbly boast a 500+ day streak. I’m still only at A2 level but am slowly getting there.

Applications & Impact

GPT is incredible for broad-based search and is just starting to be used in every day life.

However, we still believe that there are specialised, dedicated, and specific application layer companies which will be built within our two themes: Sustainable Consumption and Consumer Health. There are many use cases here including:

Byway Travel: designing 100% flight free holidays, enabled by AI
Flok Health: the UK’s first AI physiotherapist.
Sourceful: combining AI and imaging for more sustainable packaging.

We’re thinking more about AI agents within our two themes so if you’re building in either climate & health and want to connect, please reach out (estia@ekavc.com).