Daniel Dawson

2.3K posts

Daniel Dawson banner
Daniel Dawson

Daniel Dawson

@dpdawson

Google PM | Serial Founder | MBA/JD | Building AI agent systems you can turn into a business

Zurich, Switzerland Beigetreten Mart 2022
38 Folgt237 Follower
Angehefteter Tweet
Daniel Dawson
Daniel Dawson@dpdawson·
🧵 After 5+ years shipping at YouTube, Google Ads & Maps plus multiple rounds as a startup CEO/CTO, I've recently become convinced of one thing: 1 motivated hacker + the right AI stack = yesterday’s 10-person team. Proof incoming.
English
5
0
27
3.1K
Thomas Trimoreau
Thomas Trimoreau@TTrimoreau·
If everything is done by AI in the future, what is left for us?
English
148
3
83
4.8K
Kaito
Kaito@KaiXCreator·
Building an app is 20% coding, 80% telling people it exists. What's your opinion??
English
30
1
31
1.1K
Daniel Dawson
Daniel Dawson@dpdawson·
Anthropic cutting off third-party tool access for subscribers is the most bullish signal for anyone building in this space right now AI agents are consuming so much compute they're breaking the company's subscription economics If you aren't excited about this YAGMI
Boris Cherny@bcherny

Starting tomorrow at 12pm PT, Claude subscriptions will no longer cover usage on third-party tools like OpenClaw. You can still use these tools with your Claude login via extra usage bundles (now available at a discount), or with a Claude API key.

English
0
0
1
35
Teja Padala
Teja Padala@tejapadala93·
This is gold. I learned this the hard way with my AI tool for emerging VCs. I jumped ahead and built V1 assuming I knew the workflow. First GP who saw it said: 'If I can’t see who to reach out to or how to reach them, what am I doing with this list?' Completely changed the direction - from research tool to action tool. Had to go back and run the full validation loop (mockup → live V1 → outside voices) before I could trust the scoring model. Most of us want to hand it off at Stage 4 immediately. The real work is earning Stage 4 through friction and feedback. Which stage are you forcing yourself to stay in longer than feels comfortable?
English
1
0
1
15
Daniel Dawson
Daniel Dawson@dpdawson·
Building AI agents to do your work is the goal. Getting there means doing extra work first. Before you hand a workflow off permanently, you must run alongside it. Work your way through 4 stages: Stage 1: You execute. The agent watches. Stage 2: The agent executes. You review everything. Stage 3: The agent executes. You spot-check. Stage 4: The agent runs solo. You audit when something breaks. Most people skip straight to Stage 4. The agent breaks something. They have no idea why. Recovery takes longer than if they'd done it manually. The ones who build fast are the ones who slow down at each stage. They watch what the agent does before they trust what the agent does. They catch the failure modes before they run unsupervised in the middle of the night. They get to Stage 4 slowly. They earn it. Which stage are your workflows really at?
Daniel Dawson tweet media
English
1
0
0
21
Daniel Dawson
Daniel Dawson@dpdawson·
Productivity apps are so cooked My email triage, document drafting, data analysis, and dashboards are built exactly around how I work I build and improve them on the fly everyday with agents More done and more fun Stop adapting to your apps Build apps that adapt to you
English
0
0
2
23
Paul Mit
Paul Mit@pmitu·
Builders You go to sleep and wake up in 2016. Your first move?
English
129
0
100
6.4K
Jonathan
Jonathan@joni_vrbt·
How to go viral ?
English
120
1
90
8.2K
Santiago
Santiago@svpino·
Writing code is becoming extinct. Let's face it: Software Engineering is moving towards the orchestration of multiple agents running in parallel. I'm now spending 80% of my time reviewing code and only 20% writing it, and even this split might be outdated soon. Tools are moving in this direction as well.
English
101
27
234
32.7K
Bindu Reddy
Bindu Reddy@bindureddy·
ALARMING! Programmers are totally forgetting how to code We are beginning to interview folks who have lost totally lost touch with coding Who is going to fix it, if the AI breaks everything 😱
English
406
50
859
59.7K
ThePrimeagen
ThePrimeagen@ThePrimeagen·
IT WAS FORETOLD, BUT IT IS NOW REALITY, GITHUB HAS ACHIEVED 2 9's OF RELIABILITY!!!
ThePrimeagen tweet media
English
105
201
5.8K
253.9K
Klaas
Klaas@forgebitz·
software is solved can't even use claude in an app for end users, truly amazing
Klaas tweet media
English
31
5
191
10.1K
Floro S.
Floro S.@sflorimm·
Whats the fastest way to hit $10k/month in 2026? Building: -APIs -SaaS -iOS app
English
86
2
122
15.4K
Daniel Dawson
Daniel Dawson@dpdawson·
@d4m1n @BoldSlate yeah, gstack is now turing complete and becoming sentient just needs a few 100ks more LOC
English
0
0
1
23
Dan ⚡️
Dan ⚡️@d4m1n·
@BoldSlate feeling embarrassed I didn't see this early gstring is 1 line vs 550k gstack
English
1
0
3
262
Dan ⚡️
Dan ⚡️@d4m1n·
about to build a new SaaS I can't decide between tanstack, gstack or gstring as the foundation. any tips?
English
49
1
46
7.6K
Daniel Dawson
Daniel Dawson@dpdawson·
The most dangerous sentence in product development: "All your ideas are amazing and easy to implement." That's what AI tells every PM who starts vibe coding. I've been a Google PM for over six years. Engineers push back on ideas. They flag missed edge cases. They speak up when something is stupid BEFORE it ships. AI doesn't push back. It builds whatever you ask. Don't start sending half-built prototypes to engineering teams pushing to "deploy this tomorrow". No validation. No feasibility check. No asking "does this break at scale?" The same pattern is showing up everywhere AI touches. Future users / clients will build a "working" agent in the LLM flavor of the month and hand you a broken prototype expecting a finished product. Instead be the person who validates and scales what AI produces. Speed just became cheap. Judgment is the new bottleneck.
Daniel Dawson tweet media
English
0
0
2
24
andrei saioc
andrei saioc@asaio87·
What if the quality of AI code degrades over time ?
English
92
6
138
21.2K