Ben Haupt (@benchaupt) header
Ben Haupt (@benchaupt) profile picture

Building computers @synteq

Austin, TX
Joined January 2024
Photo by Ben Haupt (@benchaupt)Photo by Ben Haupt (@benchaupt)Photo by Ben Haupt (@benchaupt)Photo by Ben Haupt (@benchaupt)Photo by Ben Haupt (@benchaupt)Photo by Ben Haupt (@benchaupt)Photo by Ben Haupt (@benchaupt)Photo by Ben Haupt (@benchaupt)Photo by Ben Haupt (@benchaupt)Photo by Ben Haupt (@benchaupt)
Disclosures like this (a week before a patch) are irresponsible in the day and age of obliterated models
We are preparing a security release for Better Auth, scheduled for Tuesday, November 3, 2026. The related advisories will be published on GitHub Security Advisories at the same time. Learn more: better-auth.com/blog/securit…
9
688
Yet another hiring post! We’re looking for a systems administrator: - US Based - Experience with virtualized environments and bare metal - High growth startup - Rigorous engineering culture Dm me with why you’d be a good fit!
2
2
1
14
763
If you are; - chronically online - US based We’re hiring for growth Influencer partnerships, performance marketing, X, and DevRel. DM me with why’d you be a good fit
31
2
161
9,440
I’ve now gotten over 250 pull requests from developers trying to write the fastest possible API for a $12/month server (1vCPU 2G RAM). By popular demand, I built an official leaderboard for it: benchmarks.arjaythedev.com/ Every submission runs is tested via the following method: → Built from source on the exact same 1 vCPU box → Has to pass the full test suite before any load → Ramped up with realistic users until it breaks → Scored on the most users it holds for 5 minutes (p95 under 500 ms, p99 under 1 s, under 1% errors) The current top 3 on the $12 server: → Rust (io_uring): 264,000 users → C++ (io_uring + EASTL): 219,000 users → Hand-written x86-64 assembly: 171,000 users The range is pretty nuts. The same 1 vCPU box goes from 3,500 users (Bash) all the way to 264,000 (Rust), across 25 languages so far. The community involvement on this has been incredible. People have written their own event loops, used io_uring directly, and pushed languages I never expected to see on a server leaderboard. I’m really glad so many of you jumped in. Every implementation is open source, so you can read exactly how each one works, see how your favorite language stacks up, and send in your own. New submissions get benchmarked and added to the leaderboard automatically.
2
7
350
This Anthropic IPO will be interesting...
2
1
12
2,552
I told my QA engineer the checkout failure wasn’t a bug. He stared at me. I said, “It’s a known limitation of the current architecture.” Silence. Then: “What does that mean?” I looked at the open laptop full of red Sentry alerts. It means we have a technical debt registry. It’s a Notion page with 83 unchecked boxes, color-coded by vibes. Someone made it during a hack week in 2023. Nobody has opened it since. Including me. I told them, “We’re optimizing for long-term velocity, not short-term correctness.” They asked how I decide which failures are worth fixing. I locked eyes with them across the table and said, “You’ll just know.” They looked puzzled. I, however, understood. This is how modern engineering works. There is no severity matrix. There is only pattern recognition. And the pattern is usually something a Staff Engineer said on a podcast while holding a Yeti mug. Two sprints later another report came in. Clear reproduction steps. Video of the failure. Impacted revenue numbers attached. I asked about the root cause. They said, “Race condition in the payment service under load.” I knew right away it wasn’t the right kind of problem. Not ambitious enough. I told them we were only interested in failures that reveal deep systemic truth about distributed systems at planetary scale. They asked how many of those we’d actually shipped fixes for this year. I gave a small, knowing smile. “We’re compounding.” They asked what that meant. I said, “If you have to ask, you’re not ready.” That night I posted on LinkedIn. “Great teams don’t chase green builds. They design systems that make entire classes of bugs impossible.” 11,400 impressions. Three VCs slid into my DMs. All of them wanted to talk about “agentic reliability layers.” I marked one of the new tickets as “Won’t Fix.” I still don’t fully understand what the feature does. But the PR description said “eventually consistent multi-region consensus” five times. And something about it felt exceptional.
4
1
9
724
Come join our growing team @synteq!
Hiring a sysadmin at @synteq (cloud + GPU infrastructure). Greater Chicago or Austin area. Requirements: - Deploy HVs, remote and on-site - Have opinions on cable management - 10 years of experience with Claude Code DM me or @benchaupt. Tag someone you’d trust with prod.
3
34
2,066
We're hiring a growth ABG @synteq Must be remote / non-SF. DM me with why you’re a good fit.
33
6
149
12,733
I'm currently #1 on this hackathon, which ends on the 30th If you love me even a little bit, engage with the original tweet below so I can ensure victory (like / comment) Thanks, your favourite unc
Y'all kept asking me to share my Claude video editing skill So I built a skill maker: pick your colours, font, background, motion, and download your own in seconds Give it to Claude/GPT, and it edits in your style My entry to the @synteq hackathon 👇 videoeditskill.synteq.app/
23
4
1
82
3,076
LLMs are simply better at finding vulnerabilities than they are at writing secure code. Exploitation is fundamentally a deterministic, "loopable" problem. It is binary. Writing secure code is not. It requires consistent, non-local reasoning about invariants, threat models, and edge cases that make whole verification probabilistic and impossible. Don't forget, "fully secure" code is a fantasy.
OpenAI CEO Sam Altman: "I don't even know a dramatic enough word to use… models can find bugs that are outside of the normal threat model of what we think a hacker can do… hardware-level, insane stuff… I no longer think we're gonna be able to just patch all the bugs." It's over.
12
682
It is incredibly evident that more and more (security conscious) teams will be running frontier OS models in loops for adversarial verification and continuous red-teaming. There is simply no way around it.
1
1
8
534
If you’re a Railway customer who can’t deploy your rotated secrets right now, we’ll host your app free on @synteq while the Railway team works to restore deployments. Security has been a nightmare this week. We’re wishing the Railway team the best through this recovery & are happy to support the developer community however we can.
25
9
1
70
13,273
Incredibly proud of this launch. Developers can now push source code > deployment in one command on @synteq Cloud, with Firecracker isolation and first class secrets.
Building is fun. Deploying? Less fun. Introducing Apps and Functions, DevOps handled so you don't have to. Push to GitHub, deploy via CLI, or just drag and drop. We handle the rest. To celebrate, we're giving away $20,000 in platform credits. Comment "LAUNCH" and we'll send it your way 👀
2
3
8
546
In case anyone doesn't have a good enough excuse to build this week:
Vibecode an app and win a supercomputer. Or half a Cow 🐂 and a Flamethrower 🔥... Introducing Synteq Hack(athon). We’re giving out a DGX Spark, AirPods Max, Chipotle credits and more, plus get $20 in free credits to start building. synteq.com/hack
3
1
7
232
We’re giving away a Flamethrower & a Cow. Naturally, I’m starting with my own submission: hack.synteq.app - the platform behind the hackathon. Deployed in 1 command & 5 minutes.
Vibecode an app and win a supercomputer. Or half a Cow 🐂 and a Flamethrower 🔥... Introducing Synteq Hack(athon). We’re giving out a DGX Spark, AirPods Max, Chipotle credits and more, plus get $20 in free credits to start building. synteq.com/hack
2
1
9
432
Ben Haupt retweeted
Building is fun. Deploying? Less fun. Introducing Apps and Functions, DevOps handled so you don't have to. Push to GitHub, deploy via CLI, or just drag and drop. We handle the rest. To celebrate, we're giving away $20,000 in platform credits. Comment "LAUNCH" and we'll send it your way 👀
498
31
95
436
213,996
Only feels appropriate to ask
GoRust
50%Go
50%Rust
2 votes • Final results
1
156
Today we announced Chicago, our largest single-site expansion in company history, bringing new GPU and cloud capacity to US-Central. - 99.99% uptime - Tier III resilience - Redundant dark fiber into 350 E. Cermak
1
9
352
I have never been more pumped for a review series
I’m going to review mozzarella sticks wherever I find them and score them out of 10. The reviews will be linked below as I go. Appreciate if you send recommendations.
4
197
This ad cost $0.49 and took 22 minutes to make
🤖 Made with AI
This ad cost $7 and took 8.5 minutes to make.
1
1
5
244
(unintentionally featuring @thelukejbyrne's likeness)
1
77