AI, Software, Technology, and stuff. All opinions are my own.
Auckland, New Zealand
Joined July 2014
- Tweets639
- Following464
- Followers421
- Likes901
Our @ChatGPT plugin has been published! Now you can get UV Index forecasts from ChatGPT/dots
chatgpt.com/plugins/plugin_a…
39
Great take on the shift from chat. You're moving from tasks to outcomes.
Let the agent figure out how to get there, and evolve it's process with your feedback. Rather than you stitching up a process you have to maintain.
For me it’s a shift away from chats to roles. So my agents take on some roles. Then I give it ownership of the task, not pass it back to me.
So I have to do my expenses, my book keeper is responsible for logging into Xero every week and pulling all the receipts and setting up the transactions. To do this it needs computer use as there’s no API.
Another example is I’m building a House Hunt feature for one of my websites, where you can send properties to an estate agent you’d like to see to rent. Then I have a realtor agent find the properties and if I want, keep doing that over time. But if the app doesn’t do something I need, (like the estate agent asked if it could support comments); I just tell the realtor agent who knows my use case and intent, and it passes it to a dev agent.l to build, waits for the deployment, then carries on.
So in one way, you don’t care about context any more. They designed to be smart and avoid that. And outcomes can be linked to other actions.
Another example is I needed to do something with live chat, but my bot had all the context of policy numbers, dates and what I’m trying to achieve. I asked it to log in and drive the chat, while I put it in voice mode and watched as it interacted, allowing me to give instructions and ask questions.
It just hits different. My only take on it is that with Claude the task stops when the turns stops. Schedules are great, but they’re task driven, not outcome driven. I have to remember which project/chat the work is in.
But with Grok Bot, I direct but it owns the problem. It’s goes you would normally delegate a team to a team member. You can just say “where are we at with [thing]?”
And it feels so much quicker. Like a message app you can just keep adding context in separate messages. No stopping turns, editing, or queuing messages.
That’s my take anyway. Would love to hear if others feel the same who have used it.
1
1
43
Tech debt existed way before coding agents became a thing, and I'd argue was worse than it is now because people didn't have the time to sort it out. These days you just tell a more capable model to sort it out.
codebases have always contained slop. we have always been okay with this because code rarely needs to be perfect, outside of mission critical "hard" (eg embedded) software.
swe is a unique form of engineering because you're often building something that constantly evolves and is almost organic - user tastes change, your mental models evolve, and features come and go.
admittedly, i and many others who've been coding for a long time were initially (like back in 2024) skeptical that agents would write production code. "but what about all the slop that would enter our beautiful codebase?".
well, even before, we had human slop, and there was a lot of it! if we're being truthful, most people call "code i didn't write" slop. the reality is that 99% of people (i include myself in that number) don't even know what Good Code looks like, let alone be capable of writing it. i would propose that Good Code is:
• correct
• bug free
• fast and cost efficient to run
it has always been exceedingly difficult and expensive to write truly Good Code, because for the longest time (maybe @bendlang will change that), code was not something you could cheaply and formally guarantee to have all 3 of these properties. this is why writing hard software is much more difficult and expensive, because depending on the use case, you may only get one chance and so you have to be right from the start.
but the vast majority of software we use does not need to be Good Code to provide value to users. in fact, it is very likely that most software we use is not formally verifiable! we accepted some bugs in exchange for velocity and progress, so we rely on tests and o11y instead and have adopted a more incremental approach. this is a good thing!
agents help us move faster, and if you're open to the idea, are capable of helping us write better code. it won't be Good Code (that is still expensive), but it is likely to be much better on average at scale.
the new craft of swe is to figure out how to do this. it's your job now! if your productivity with and without agents is not significantly different, you're doing something wrong. but it's not too late to embrace it and use it as an opportunity to learn a whole new skill tree
1
34
Is the play just to give @bot a logged in session of Claude Code so it can delegate all the building to it and be the orchestrator?
Love the doing part of bots but I still have a strong affinity to @ClaudeDevs for the building.
1
1
69
If you want your @bot to make calls, just point it to ringforme.com if you live in Australia/NZ. At least until @poteto builds it 😅
1
1
56
I have come to the realization that Dots is a huge unlock if you have a lots of things to juggle at once in your work or life. Otherwise you probably won't feel the benefits as you could just go and do those things already in chat/code.
Did I get that right @TheRohanVarma?
1
94
Sometimes, you just want to paste things into another computer. Or have a shared clipboard with your agent.
Made with Opus 5.5, including the video @ClaudeDevs
2
2
142
Hot take: if you don't notice the difference between AI models, you're not pushing them hard enough.
1
60
Prediction: @OpenAI is going to launch outbound calling for ChatGPT.
In the meantime, I used Astra to build a working product using GPT-Live-1 which solves outbound phone calls and has MCP.
Working really well and saved me a few calls already!
ringforme.com
1
1
113
There's a lot of talk about agentic loops. It makes sense because if you think about it, work is just a daily/monthly/yearly loop anyway.
47
I wonder how many people especially in FinOps landed in the hospital because of this...
📣 Service Update: AWS Billing Console
We are investigating issues with Cost Explorer reflecting inaccurate estimated billing data.
📊 Track status: go.aws/aws-hd
#AWS #AWSSupport
101
