Teren Perera

Founder @ Papercut Labs · seeing what happens if we make ai do literally everything · pushing claude code to the limit

[Teren Perera @terenpcl]

@kunchenguid anything human facing in the flesh (and maybe where voice is involved but increasingly less so) ie obviously dentist but receptionist, procurement etc. needs a human at the interface layer

if above is true, AI every day and twice on Sunday

otherwise human probably

[Teren Perera @terenpcl]

@kunchenguid I would imagine that the value comes from regression testing vs. the first build

[Teren Perera @terenpcl]

@kunchenguid this is why all roads lead back to the cloud

[Teren Perera @terenpcl]

@quxiaoyin ie take your knowledge of how to run an agent-first org and go start something of your own where you can make the decisions that the bigger companies are too risk-averse or have too much internal friction to make

[Xiaoyin Qu @quxiaoyin]

I am gonna say something risky:
Most of the FDEs I've seen are solving BS problems.

1. Existing enterprise workflows were designed around people, not agents.
2. Real AI gains require rebuilding the org chart.
3. Most execs have no incentive to do that, and most CEOs aren't in founder mode.
4. So FDEs automate the status quo, because that's what's easy to sell. But the status quo is stupid and will be obsolete.
5. If you FDE on critical workflows, 1) enterprises are really this dumb to hand off trade secrets? 2) if you know that, shouldn't you start a competing firm?
If you FDE on non-essential AND soon-obselete workflows, did you learn anything meaningful and repeatable?
6. If you are a smart young FDE, why spend your best years automating dying workflows and managing people who fear AI will take their jobs? Shouldn't you start an agent-native firm from scratch in that industry instead?

Analogy would be, instead of teaching everyone to be AI builders within a tech company, you as FDE build agents that auto-ingest meeting notes for PM, QA, front-end, back-end, devops, UX and help each of them write PRD, Figma, and meeting notes faster for their sync meeting, but the right solution should be to shrink roles, hire full-stack AI builders, and kill meetings. You gain lots of experience with your PRD/Figma/standup meeting sync agents but it's a useless skill.

I am not a FDE expert so I might oversimplify things. I am genuinely curious about this topic and welcome any critiques and comments.

[Teren Perera @terenpcl]

@kunchenguid which, ironically, having a big brand helps with marketing and consumer trust, and the bar for that for something this intrusive I think is quite high outside the geek community

which would be the pessimistic outlook for what you're describing

[Teren Perera @terenpcl]

@kunchenguid does that mean in your opinion hermes/oc are one (big, prohably) ux/onboarding away from achieving your list?

oc still needs to have the infra layer but hermes has it with cloud

[Teren Perera @terenpcl]

@thsottiaux as opposed to what exactly that was happening before

[Teren Perera @terenpcl]

@janwilmake interesting can I ask if you are experiencing any downsides with this approach

[Teren Perera @terenpcl]

@kunchenguid I run a m3 ultra studio but I've found that at its possible to heavily stress it esp when many agents do big file operations

I keep work on the studio until a certain load then spin up ephemeral https://t.co/tDBbL09rVh machines which charge by the minute to take on excess load

[Teren Perera @terenpcl]

@signulll

[signüll @signulll]

openai responding at length to anthropic’s ads was a huge pr self own imho. i’m truly neutral here, so i can break it down. i just like good competiton.

1) this reads like it was assembled in a war room by committee. you can smell the post it notes. the “more texans use chatgpt than claude” line is especially bad.. someone clearly thought it was clever, then couldn’t let it go, so it got reused multiple times. sounds like insecurity more than confidence.

2) why on earth would you expect a competitor especially one with a totally different ideology to engage with your internal narrative? that’s not how ads work. anthropic wasn’t speaking to openai. they were speaking to users of chatgpt. responding as if this is a good faith philosophical debate is a category error.

3) the optimal response was likely silence. the second best response was a single graph of active users with the caption “lol”. anything more is just validating the frame you’re supposed to ignore.

by responding seriously, openai elevated the ads, accepted anthropic’s premise, & turned a joke into real discourse. this is usually a classic mistake of people who are powerful but insufficiently online.

they should’ve consulted someone who lives on the internet not someone who lives in talking points.

[Teren Perera @terenpcl]

@ksaitor that's the 11th PR

[Teren Perera @terenpcl]

@ksaitor https://t.co/I9xHJDhvay

I think a) yes, but b) bringing back the dinosaur counts as value adding cause of the impact it's going to have

[Teren Perera @terenpcl]

this looks very cool but I think it's a testiment to ui/ux more than a technical breakthrough. it's was possible (at least on Claude Code) before. basically it's just the connector returning a url that can be rendered in an frame, so you could imagine that a coding agent could open up the browser on receipt of such a url. but there a bunch of powerful benefits to this being released:

a) no previous ability to display such a return on Web or mobile based AI clients. trivial technically to implement conditional frames but the fact that they can now unlocks an entirely new paradigm of interaction

b) just because it was possible before didn't mean that anyone was doing it, making it a standard capability is going to get a whole bunch of companies thinking about this mode of UI with users

[Teren Perera @terenpcl]

this looks very cool but I think it's a testiment to ui/ux more than a technical breakthrough. it's was possible (at least on Claude Code) before. basically it's just the connector returning a url that can be rendered in an frame, so you could imagine that a coding agent could open up the browser on receipt of such a url. but there a bunch of powerful benefits to this being released:

a) no previous ability to display such a return on Web or mobile based AI clients. trivial technically to implement conditional frames but the fact that they can now unlocks an entirely new paradigm of interaction

b) just because it was possible before didn't mean that anyone was doing it, making it a standard capability is going to get a whole bunch of companies thinking about this mode of UI with users

[Alex Albert @alexalbert__]

We’ve launched the first official extension to MCP.

MCP Apps lets tools return interactive interfaces instead of just plain text.

Live in Claude today across a range of tools.

[Teren Perera @terenpcl]

@SherryYanJiang in fact, the opposite way works a lot better - eg. 'ask me some questions to pick something to ____' learn, improve on, whatever

with infinite possibilities, humans get stuck. very underrated use of AI is to narrow down the possibilities

[loaf @lordOfAFew]

This is a moment in time.

Openai will be close behind. Followed by Google.

Microsoft will do something similar and distribute to billions instantaneously.

Apple will incorporate a similar technique

And like that the entire world economy will shift over 12 months.

The promise of what has been told to you for the past 3 years is here - and it’s this ergonomics. Agents on your Mac/pc, operating on files like how you work.

After this comes billions of agent transactions accelerating information transfer in the purest form of capitalism, accelerating civilisation.

What a time to be alive. Stop scrolling, start building the future.

[Teren Perera @terenpcl]

@badlogicgames idk if copy is the word

I bet that 'Claude code for non-technical people' has been sitting on a whiteboard for months. the fact is that many of these ideas are inevitable and many people are converging on it.

general tooling is going to be taken by the big players, find a niche

[signüll @signulll]

this is genuinely good. like quietly devastating good.

how did anthropic beat chatgpt to this?