@scotups

2x @ycombinator founder, prev Investor @Initialized, average engineer @netflix

 San Francisco, CA
Joined May 2013
  • Tweets3.1K
  • Following937
  • Followers14.5K
196Media
Most AI demos stop at a chat box. Real products need background agents that process jobs, react to events, and retry failures long after the user closes the tab. 🤖 Join Netflix Senior Engineer @scotups on September 22 for a live, full-day workshop: Build AI Agents that Never Sleep. Learn production patterns for Node.js agents, including queues, durable state, recovery, guardrails, and observability. You'll leave with a complete hands-on reference architecture. 🗓️ September 22, 2026 at 9:30 AM - 4:30 PM CDT RSVP here to level up your AI engineering skills: master.dev/workshops/backgro…
2
33
4,451
Unless you’re doing something leisurely like gaming or just love the nostalgia of it, soon you won’t be using computers. They’ll be almost entirely replaced by AI and become tools AI use themselves.
2
1
5
1,116
Replying to @scotups
@scotups, someone known for being heavy in AI, just described his dev workflow as "pretty boring." No parallel instances. No chaos. Just a fast editor, Claude Code in the terminal, and spec-driven development. Full interview: youtu.be/MYLsBdifOUg?si=u3S0…
1
7
35
2,398
Is your agent great for demos but not production ready? Join us on June 12th, as @scotups (senior engineer at Netflix) teaches you the infrastructure layer that fixes that: durable execution, sandboxing, multi-agent orchestration, HITL, and more! 👉 master.dev/workshops/harness…
1
3
19
2,087
70% of AI Engineer job postings in 2026 center on RAG, evals, agents & production deployment. Learn with @scotups in this 2-day workshop. 📅 April 6–7 | 9:30am–4:30pm CDT | Live Online frontendmasters.com/workshop…
3
64
3,924
Damn, didn’t even know this was possible. Love it
kev
@PhatStrawsMar 28
I built clawtv — a CLI that gives Claude Code full access to my TV. Tell it “Open Netflix, find Inception, pause on the play button” while I’m out. It screenshots the screen, navigates via local ADB. No cloud. No API keys. I walk in and it’s already queued. Pure magic.
3
7
1,833
AI data centers have lost the plot. Most business use cases don’t need large frontier models. If you’re running inference around the clock the unit economics don’t make sense for cloud. Keep your eyes on Apple. The M5 is a turning point.
10
887
Don’t make hard work become your whole personality. You’re not missing out on anything. This token anxiety issue is really sad imo. I can’t tell you how many times as a VC I gave advice to founders to stop writing code, that’s still true today even if it’s Claude generating it.
1
15
1,306
There are millions of apps (web, native, unfinished in github) that have 0 users and generate no value. Code has never been the bottleneck and today's AI does not change that.
3
2
24
1,450
A really good coding agent is really good at saving a really good engineer’s energy. If your company is seeing more value because all your engineers are using a coding agent, then I’d argue you did a subpar job on hiring
1
1
5
607
Also, although we can create more code than ever before, software is still the same today at best as it was pre gpt 3.5. So are user expectations. You can only create so much “value” in today’s software and user expectations.
2
514
LLMs aren’t MOATS. Interfaces are. UX/UI designers, 2026 is your year!
2
1
15
949
My opinion is drinking the 996 kool-aid won’t save your startup. It’s not going to be the thing that makes you win. If you’re good at building a product that sales well to a hefty market then I’d argue you’d be working less, especially given how free and cheap automation is
1
2
1,140
I’ll share a small part of pickle.com Back in med school, I became obsessed with augmenting memory and dreamed of a Notion or Obsidian that completes itself. Today, we’ve built something close. My self-awareness is sharper and everything feels connected. I genuinely believe AI does not replace humans. It amplifies us. Huge respect to our engineers and designers who made this crazy thing real. Bubbles are the episodic units of my life that the system interprets from my raw data. Clouds are the system’s questions, its hypotheses about who I am. When I answer a cloud, it becomes a bubble again. There is so much personal data that I cannot fully demo it. Wish I could. This system understands me more deeply than anyone. Want to try it? Retweet and comment “memory.” I’ll DM you an access code to skip the waitlist.
1,564
905
171
4,411
660,238
Traditional VC model is going to look very different in 2026. Many over funded startups will die unfortunately and the reaction to this from founders and firms will be drastic. Capital is till needed, the vehicle and the discovery is what’s broken.
3
1,259
MCP is a barrier for AI. You don’t need MCP servers to hit that API, just let your LLMs build its own tools.
3
2
11
2,296
This aged very well: github.com/universal-tool-ca… along with other research seems to agree with my observation

GitHub - universal-tool-calling-protocol/code-mode: 🔌 Plug-and-play library to enable agents to...

🔌 Plug-and-play library to enable agents to call MCP and UTCP tools via code execution. - universal-tool-calling-protocol/code-mode

github.com
1
3
739
Productivity might be the least lucrative and interesting category to point gen AI at, and yet the whole industry is stuck there.
5
1,234
Assuming AI is benign or aligned, one huge problem we’ll face as society is the gap between AI being good enough to put folks out of jobs and not quite good enough to make up for it yet (scientific breakthroughs)
1
3
77
16,840
Well yea, the point of it having to exist in the first place and who created it is the issue I believe. Depending on how many tools / servers you have and the task you’re doing, could be less or more tokens. Just my observation lately.
3
1,038
Load more