Having an agent edit 100 files means the job will definitely get done correctly. When it writes a script to bulk edit things it fucks up and spends ages debugging their script.
> LLMs are a great foundation but building your own harness is a huge undertaking, a deep rabbit hole
I’ve been doing this for the past few months. I started with a server where I ran pi in tmux and then used that to build an LLM gateway and agent session manager, then built deterministic workflows using bash scripts and a skill/script distribution system. The app works on desktop, mobile and web and it works great. Non technical colleagues are using it to build and ship real software and it’s cheap AF even using API pricing because it works well with Luna or deepseek.
This is the "i made a voice controlled agent" thread all over again. lol, I too have made a stable of harnesses and tools to run them and have different levels of them monitoring each other and different spends to code/review/triage.
I also use Luna and DeepSeek in a custom harness. (Mine is very minimal.) GLM also works great.
I had issues with some other models but it seems to do with the system prompt and tool calling format. Some models seem to only work well with some harnesses.
Ha yeah GLM and Kimi seem to always biff the tool calls. I recently added in a raw stream log switch to diagnose but since DS and Luna are so good and cheap I haven’t been that motivated to debug it …
I wonder why not Kotlin Multiplatform. I’ve been using LLMs to build with KMP for about 10 months and it’s a delight and you can use native Swift UI when you want to. It seems to be the best of all worlds.
Well the thing is that the server code is basically Java, and the only bad thing about Java is writing it but since I don’t have to write it, it’s actually a really solid option. Then you have shared UI (don’t use WASM for web though it’s terrible, I use React for the web frontend) but you can switch to some or all native components when you need to on any platform. So for example I needed native components for secure storage and passkeys and did that by mixing in some swift on iOS and using a JNI native bridge on macOS and Windows.
I created a standardised bootstrap for all our projects here starting from the KMP wizard:
The reason it the bootstrap has all platforms included is that adding in iOS, android and desktop after you have already created the project is a bit of a hassle compared with just always having the same structure even if you only want to use web.
Then you can easily add on native apps later if you like.
I love how the marketing departments of these things always imagine people are just constantly plagued by the hassle of organising their flights, restaurant bookings and movie tickets
I liked how in the example here it tells the user "Your flight is delayed 2hrs due to storms in Denver..."
Then says "checking alternate routes" then shows them a card with their original flight details (it does have "Delayed" in red at the top) and asks them if they would like to change their flight.
The user responds with "Yes please! :thumbsup:"
The AI does not provide any information about the new flight, and only displays a card for the original flight, which being the original, is the day and time the User likely wanted to travel.
Between it being the only piece of information on the screen and the status of "checking for alternate routes" it will certainly be a surprise for the User when their Sept. 17th 5:45 departure has been rescheduled for Sept. 19th to avoid a two hour delay. Helpfully replaced with a 2 day delay, thanks Muse!
Most airlines let you change flights, usually for a fee plus the difference in fares. The fee may be zero if your original ticket allows it -- sometimes called "flexible fare."
It's because people who create these use cases perceive this as a difficult and painful task - but how many people really think about their flights over the day? Pretty sure the majority of Meta users has not flight booked for the next 3 months. I can only assume that this is more seen as a showstopper. Look how difficult this is and how we solved it.
Shopping is another classic. I personally don't want to trust an AI agent with checkout. research is fine, but I don't want the AI to go on a shopping spree or Amazon added some subscription into their checkout flow and suddenly, I'm paying for Prime because my AI agent thought it's a good idea or just missed it.
It's not unheard of that Amazon bundles a subscription into your checkout. So I don't think it's too unrealistic.
Not to mention the adversarial risk. Prompt injection, LLM-maximized advertising, and just common scams are going to take so much money if these catch on.
handling end-to-end flight booking seems like a very good target use-case that if you can get right, implies that you got a lot of other things right. Just like how Amazon started off by selling books online.
And your emails! It can manage your email and schedule! Wow, so great, it can handle all the 10 DHL notifications and other spammy things I got over the past week. And my empty personal calendar
The original selling point of mobile web (tiny screen with text-only data, before real mobile web) on mobile phones or even watches was checking stock prices and weather. It was really weird that checking stock prices was something you need to do on the go, especially from the watch.
The average person is not using a calendar and is not getting more than 1 relevant personal email per week.
Most people use their personal email extremely sparingly and it is inundated with promotional / account-related materials which are already handled rather deftly by most mainstream email providers
The average kid in an average suburban public school is not getting this amount of digital communication. If they are, they still aren't using a Google Calendar and are managing the complexity in much simpler ways (paper calendars)
I think they mean the parents.
I got 5 emails from the kids school today, and have regular calendar events related to the kids.
So yeah this is for parents
To be fair I do get fucking hammered by school emails, and not just school school but dance school.
But like, an AI would just tell me what’s in the email and ask me what I want to do about it which is almost always ignore it and I know what’s in the email because I have eyes.
lol no one is getting personal emails, maybe some people get one every few months, but most probably never (not taking into account other than order/shipping confirmations, login codes, etc)
First of all, the average parent is definitely doing some volunteering, at least in the US. From rec sports to scouts to church stuff to local committees, it's very rare for someone to be totally checked-out on all those counts.
And as a bit of anectata: I play two sports and my child plays multiple. Every single one of those groups or teams relies on email, even when other modes of communication are layered on.
I’m pretty sure if you asked my mom what problems AI could help solve for her, navigating website’s purchasing/account flows and directly answering questions about the contents of her email would be #1 and #2.
And my mom doesn’t use AI products like chatgpt because she’s not a student, nor agents because she doesn’t work in tech. To her AI, is no different from those chatbots websites pop up in the corner, the ones you never seek out or interact with intentionally, because you don’t need what they offer.
So I actually think this kind of marketing is quite helpful for regular people who mostly just use their phones to buy stuff, looking things up, navigate life, and entertainment. My mom doesn’t give a shit about APIs or sandboxing, or benchmarks and to her it literally is a hassle to manage a million one-off accounts and confirmation codes and receipts when she just wants to buy something on her phone. She would not assume AI is capable of that or know how to set it up locally.
that’s not even getting into the fact that the top 10% controls 50% of consumer spending in the US and primarily purchases convenience + health + experiences, whereas the other 90% primarily prefers to buy aspirational/identity based goods that evoke their mental model of a higher status lifestyle. It’s why the same bustling lifestyle archetype you criticize is literally used exclusively in car ads, housing marketing materials, consumer electronics, home goods, etc. Because normal people want to feel and/or look good not spawn subagents
i've gone on many trips before, and one of the most fun parts of it is doing the actual planning and finding stuff
i'd much rather AI do the boring shit, like when it comes to insurance with multiple parties and service providers involved that don't talk to each other at all, making things 10x more complicated than they need to be
If you want to know why Silicon Valley Execs think these are such huge pain points, think about what Silicon Valley Execs farm out to their armies of personal assistants. Some Senior Vice President guy probably has a lot of restaurant bookings and would forget them if they didn't have a PA to help organize them.
AI itself is a product that could only come from the mind of a corporate executive: It's a cheerful, helpful sycophant, who always does what it's told and stays in its lane, has infinite energy and memory, and allows the user to concentrate on his golf game instead of all those mundane other things. It's the perfect employee, from their perspective.
Or in Muse’s case, helping your child pick out school supplies for their third grade year.
Meta’s ideal parent is too busy to spend time with their kids, so they let the AI pick sponsored garbage for their cart and check out with their Link payment details.
i think product just uses these use cases as their example demos and then when they show it to marketing theyre like "OK I guess this is what it's for".
Yeah it’s pretty new but they’ll get there. I was using IntelliJ and basically wasn’t using it as an IDE. Then I tried Air and I was like “they get it”. But yeah they have work to do to get it out of preview. They are very responsive though
When I was bringing up IDEs at work and advocating JetBrains (I’ve since built my own … ) a colleague questioned why we don’t just use cursor instead of waiting for JetBrains Air to come out.
The products are basically equivalent but JetBrains is a profitable bootstrapped company that has been building IDEs for 20 years and Cursor is a flash in the pan startup with a squillion dollar valuation to support. Sooner or later they’re going to do something that will fuck their users over.
reply