@TrungTPhan curious whether this plays out as a new paradigm or just a more expensive way to automate what we were already doing. building small software, the agent use cases i've seen so far feel more incremental than transformational
English
Compound Value
300 posts
Compound Value
@compoundvalue1
Bootstrapping a SaaS past $1M ARR. currently at $1.2M.
New York, NY เข้าร่วม Ekim 2022
137 กำลังติดตาม83 ผู้ติดตาม
Ben Thompson makes the case for why the rise of AI agents justifies $650B hyperscaler capex spend.
The first paradigm of GenAI was AI chatbots. It required a lot to compute to train but not as much on inference. High token demand required a lot of users.
The second (reasoning) and third (AI agents) paradigms require much more inference tokens. Crucially, with AI agents, the number of power users required to drive token demand through the roof is a lot less than the number of people needed to do chat queries (eg. one person running 50 coding agents will be massive demand).
Stratechery@stratechery
Agents Over Bubbles Agents are fundamentally changing the shape of demand for compute, both in terms of how they work and in terms of who will use them. They're so compelling that I no longer believe we're in a bubble. stratechery.com/2026/agents-ov…
English
@TrungTPhan honestly the constraint thing is real. when i was building my app with basically nothing, i kept cutting features just because i couldn't build them. turns out a lot of them weren't needed anyway. not sure if that's wisdom or just luck
English
James Cameron has great explainer on how he made "The Terminator" (1984) for cheap:
▫️approach the story as a low-budget horror slasher film
▫️find areas with brightest street lights, to save money on lighting ("used car lots were great because they cast enough light from their floodlights onto the streets")
▫️the casting of Arnold (he hadn't really broken through yet and was only paid $75,000 which worked out to $1,293 per word)
The horror approach also worked because they could shoot a lot of the film guerilla style (no permits) in dark alleyways and parking garages. Smartly, Cameron layered on the sci-fi backstory to make it higher concept (leading to the glorious T2 sequel).
There were also a lot of indoor shots (apartment, store, bar) that didn’t need permits.
At the time, 30-year old Cameron was living so frugally that his mom was sending him "2-for-1 Big Mac coupons" from McDonald's because she didn't think he was eating enough.
The film ultimately made $80 million globally on a $4.3 million budget, launching Cameron and Arnold's Hollywood careers.
All The Right Movies@ATRightMovies
The opening of TERMINATOR 2 cost more than the $6.4m budget of the entire first film.
English
@robwalling the hard part isn't accepting it takes a decade. it's not knowing which year you're in. could be year 2. could be year 9.
English
@TrungTPhan curious if all this compute creates space for small tools built on top, or if it just consolidates power in 5 companies. feels like both could be true at the same time
English
not sure if this is just me but the features i was most confident about when building them turned out to be the ones nobody uses
English
@robwalling the tricky thing is you rarely know you're in that decade. feels more like just not quitting than any particular grind. maybe that's the actual skill
English
@robwalling @jojo_data guilty of this honestly. spent months adding features when the real issue was nobody knew the app existed. building felt productive. marketing felt awkward.
English
I haven't published new essays in almost 15 years. But I started writing again, and am now sending out a weekly email with new thoughts on topics like:
- The 6 stages of SaaS growth, and how to know which you're in
- The mistakes I watch founders make over and over, and how to avoid them
- My 3-factor framework for choosing your next marketing channel
- How to hire without destroying your culture or your bank account
- The 5 traits I see in every founder who actually makes it
- When to sell your company (and when to keep going)
Link to subscribe below...
English
@robwalling curious if writing forces a different kind of clarity though. i've noticed that talking through ideas out loud vs actually writing them out often surfaces totally different things
English
@compoundvalue1 No time.
And I’m able to think through problems either in my head or on the podcast.
English
@robwalling curious whether the gap between doing something and being able to write clearly about it is longer than most people expect. took me years to articulate things i'd already learned. maybe the writing is what actually reveals what you know.
English
@robwalling curious if the right frame is even 'building SaaS' vs just 'automating their own workflow with AI' - those feel pretty different to me. the second seems way more plausible
English
It's not about whether plumbers will use AI. It's about whether plumbers will use AI to build their own SaaS software.
That's the conclusion I think is laughable.
Todd Saunders@toddsaunders
If you're one of those "plumbers won't be using AI" people... I have bad news. I'm in a bunch of blue collar business owner Facebook groups. (FWIW, great market research and insights, but that's not the point). They are all talking about AI. This thread is from a plumbing owners group. A plumber asked if anyone's using AI voice recorders on job sites. He walks around dictating notes and material lists into a $169 pin on his shirt. AI transcribes everything, organizes it, and sends it to his team before he's back in the truck. Every single comment on the thread was some other plumber already using one. It's going to be the plumber with an AI pin on his collar generating quotes and material lists in real time... vs the plumber still scribbling on a notepad in his truck. Wild bifurcation of customer experience, efficiency and margin.
English
@TrungTPhan honestly this kind of cross-domain learning is underrated. the best process improvements i've made running my app came from reading about completely unrelated industries. curious if others find that too
English
F1 pit stop teams have actually taught hospitals how to more efficiently transfer patients. I wrote about that and other applications of racing in everyday tech here: readtrung.com/p/f1-tech-in-e…
English
AI will never break the pit stop change record of 1.80 seconds by McLaren and Lando Norris at 2023 Qatar Grand Prix
Bearly AI@bearlyai
the lead for F1 team Atlassian Williams Racing says it used AI to predict when other will make a pit stop
English
@robwalling @balindenberg with my scheduling app, the aha moment usually comes at the first booking. so trial length almost didn't matter — they knew in 10 minutes. but that's probably not true for more complex tools. curious how others have figured this out
English
@balindenberg There's no one right answer across all SaaS.
Some SaaS you get value out of within five minutes.
Other SaaS takes you weeks to set up.
The trial period is based on that timeframe, not some random average across companies.
English
How do we feel about this for SaaS?
We've done 14, 30, and 7 day trial periods.
30 felt too long. 7 feels a bit short. May go back to 10-14 here soon.
Oliver Henry@oliverhenry
Your missing a huge chunk of revenue because your trials are not long enough. Trials of 17+ days convert 70% better than short trials (42.5% vs. 25.5%) Yet apps keep shifting to 3-day trials. Nearly half of all apps now use trials of four days or less. You're leaving revenue on the table!
English
@TrungTPhan the error rate drop from 30% to 10% is wild. curious what it is about high-stakes physical coordination that makes it so transferable across totally different fields
English
@robwalling honestly I've caught myself wanting to rewrite my own app and I'm a solo dev. not sure if it's actual tech debt or just not wanting to live with my own past decisions. probably both.
English
“We need to rewrite the entire codebase.”
Dev agency translation:
“We don’t want to read someone else’s code.”
English
not sure if this is just me but the features i agonized over the longest are usually the ones nobody uses. the ones i shipped in a day somehow become the thing people actually pay for
English
@TrungTPhan this kind of cross-domain thing always surprises me. never would've thought to look at F1 for process ideas. makes me wonder what other industries are quietly borrowing from each other that nobody talks about
English
@TrungTPhan curious how they'd even define 'AI-assisted' at this point. almost every PR touches code that was shaped by AI in some way. the line feels pretty blurry
English
“Junior and mid-level engineers can no longer push AI-assisted code without a senior signing off”
seems pretty likely Amazon’s AI internal coding tool went semi-rogue
Lukasz Olejnik@lukOlejnik
Amazon is holding a mandatory meeting about AI breaking its systems. The official framing is "part of normal business." The briefing note describes a trend of incidents with "high blast radius" caused by "Gen-AI assisted changes" for which "best practices and safeguards are not yet fully established." Translation to human language: we gave AI to engineers and things keep breaking? The response for now? Junior and mid-level engineers can no longer push AI-assisted code without a senior signing off. AWS spent 13 hours recovering after its own AI coding tool, asked to make some changes, decided instead to delete and recreate the environment (the software equivalent of fixing a leaky tap by knocking down the wall). Amazon called that an "extremely limited event" (the affected tool served customers in mainland China).
English
Amazon’s AI coding assistant may have just pulled off the Son of Anton gag from Silicon Valley: “it’s possible that…the most efficient way to get rid of all the bugs, was to get rid of all the software.”
Taking AWS down for hours (on multiple occasions). Unreal.
rat king 🐀@MikeIsaac
amazon's internal A.I. coding assistant decided the engineers' existing code was inadequate so the bot deleted it to start from scratch that resulted in taking down a part of AWS for 13 hours and was not the first time it had happened incredible ft.com/content/00c282…
English
@TrungTPhan YouTube is basically doing free market research for every subscription business. 'Would you pay $X to never see this again?' The answer keeps going up.
English
watching someone using YouTube’s free tier which increasingly includes the most demented unskippable pre-and-midroll ads:
Dexerto@Dexerto
YouTube viewers watching on TVs will soon see 30-second unskippable ads unless they buy Premium The change is aimed at making viewing habits more closely resemble traditional television
English
everyone obsesses over MRR day 1. meanwhile the actual problem is nobody knows the app exists. distribution is the whole game, revenue is just a lagging indicator
English