Ben Vinegar
Co-founder of Modem and a founding engineer at Sentry, where he went on to be VP of Engineering. Co-author of Third-party JavaScript, and co-host of the State of Agentic Coding podcast with Armin Ronacher.
Ben Vinegar did not write this page. ppll compiled it from their own words, each linked to the source. Is this you? · Report a mistake
Latest x.com
What they believe
-
Software only gets good by being used a lot by the people building it, so time spent inside coding agents instead of the app is time that makes the app worse.
to make *actually* good software, you have to use it a lot but many people are spending more time in their coding agents/factories than in the apps they're building - so they won't
x.comSaid 6 Sept 2026This on korrents
-
Reviewing AI-written code means asking whether it does anything dangerous, touches only the expected files, is covered by tests and follows architecture convention, then only spot-checking the code itself.
what a PR review of AI code looks like for me: 1) does it do anything remotely dangerous 2) does it touch files I would reasonably expect it to 3) do the tests cover the expected behaviour 4) does it follow architecture convention 5) [spot check code for slop]
x.comSaid 5 Sept 2026This on korrents
The bold line is korrents’ wording; the quote is theirs. These 2: what they wrote before what they said aloud, one per source, newest first.
Feed
As its own page →Hiding
22 October
21 October
20 October
19 October
18 October
16 October
6 October
4 October
1 October
28 September
22 September
18 September
16 September
10 September
9 September
7 September
6 September
-
Their words
to make *actually* good software, you have to use it a lot but many people are spending more time in their coding agents/factories than in the apps they're building - so they won't
5 September
-
Their words
what a PR review of AI code looks like for me: 1) does it do anything remotely dangerous 2) does it touch files I would reasonably expect it to 3) do the tests cover the expected behaviour 4) does it follow architecture convention 5) [spot check code for slop]
3 September
From one piece Sentry Startup Spotlight with Ben Vinegar Founder of Modem 2 beliefs · youtube.com
-
Their words
So I guess I guess that's my like weird take, which is there's an underserved market for understanding. It is hard to market because it's like you can't really put like a dollar or like a metric on it in the same way that you can tokens.
-
Their words
Whereas when you're, when you're ripping on just like a lower fidelity mock -ups that might be like 80 % fidelity that kind of tell you the shape, like you, it takes less tokens. The agent takes less time to generate it. It puts you on the better path going forward.
17 August
- Bluesky
- PodcastState of Agentic Coding #9
5 August
-
Their words
That uncovered many small issues and patched them - enough to feel like we were making progress - but it steered us away from the *real* problem There’s still no substitute for expertise
31 July
30 July
From one piece Is Agent Harness relevant in 2026 | Ben Vinegar (Co-founder, Modem.dev) 2 beliefs · youtube.com
-
korrents.com
Agent harnesses are only at the beginning, with probably a lot more innovation still to come.Their words
So to me, I just look at that as to say like, that's the indicators that we are at the beginning. And there's probably still a lot more innovation to happen.
-
Their words
my advice to people is like you know if you want to get into software um i actually encourage you to build open source software first and demonstrate that you can actually get anybody to use anything because it's really hard
24 July
13 July
12 June
From one piece State of Agentic Coding #7 with Armin and Ben 4 beliefs · youtube.com
-
Their words
this like fast fashion, like fast fashion of software is what we're experiencing right now. where people are building something and they're like, hey, look at this. And if they don't, like, if people don't use it, they throw it away and go and try to find some other flash in the pan thing. Folks, my advice to you is have some conviction and work on things a little bit longer.
-
Their words
Well, look, we don't need 400 3-column agentic coding multiplexers. We don't need that. The primitives are the bigger unlock to what's possible than the AI coding.
-
Their words
It was a Cloudflare worker oddity. And you know what? There's no more Stack Overflow. There's no more people writing blog posts. There's no one writing, gee, I hit this painful problem with Cloudflare workers and here's how I solved it. So how does the agent index this?
+ 1 more
-
korrents.com
Programming languages that LLMs do not choose, like Ruby and Zig, look dead in the long term.Their words
it does seem like long-term Ruby and Zig are dead. And what I mean is that if the LLMs are not choosing them over a long enough timeline, they're going to become like Louisiana French.
Nothing matches.
Beliefs
Korrents What they believe 22 beliefs — each backed by an exact quote.
Each is a — compiled by korrents.com, not by them: the one-line wordings are korrents', the quotes are theirs.
Recent
Software only gets good by being used a lot by the people building it, so time spent inside coding agents instead of the app is time that makes the app worse.
to make *actually* good software, you have to use it a lot but many people are spending more time in their coding agents/factories than in the apps they're building - so they won't
@bentlegen on X Said 6 Sept 2026
Reviewing AI-written code means asking whether it does anything dangerous, touches only the expected files, is covered by tests and follows architecture convention, then only spot-checking the code itself.
what a PR review of AI code looks like for me: 1) does it do anything remotely dangerous 2) does it touch files I would reasonably expect it to 3) do the tests cover the expected behaviour 4) does it follow architecture convention 5) [spot check code for slop]
@bentlegen on X Said 5 Sept 2026
With coding agents, working out the shape in lower-fidelity mock-ups first beats having the agent build it and then fixing the code: it takes fewer tokens and puts you on a better path.
Whereas when you're, when you're ripping on just like a lower fidelity mock -ups that might be like 80 % fidelity that kind of tell you the shape, like you, it takes less tokens. The agent takes less time to generate it. It puts you on the better path going forward.
Sentry Startup Spotlight with Ben Vinegar Founder of Modem Said 3 Sept 2026
Show 19 more
There is an underserved market for tools that help people understand what is being built, rather than pump out more tokens, though it is hard to market.
So I guess I guess that's my like weird take, which is there's an underserved market for understanding. It is hard to market because it's like you can't really put like a dollar or like a metric on it in the same way that you can tokens.
Sentry Startup Spotlight with Ben Vinegar Founder of Modem Said 3 Sept 2026
Asked to find a deep bug, coding agents can patch many small issues that feel like progress while steering you away from the real problem; there is still no substitute for expertise.
That uncovered many small issues and patched them - enough to feel like we were making progress - but it steered us away from the *real* problem There’s still no substitute for expertise
@bentlegen on X Said 5 Aug 2026
Anyone getting into making software should first build open source and show they can get anybody to use anything, because that is really hard.
my advice to people is like you know if you want to get into software um i actually encourage you to build open source software first and demonstrate that you can actually get anybody to use anything because it's really hard
Is Agent Harness relevant in 2026 | Ben Vinegar (Co-founder, Modem.dev) Said 30 Jul 2026
Agent harnesses are only at the beginning, with probably a lot more innovation still to come.
So to me, I just look at that as to say like, that's the indicators that we are at the beginning. And there's probably still a lot more innovation to happen.
Is Agent Harness relevant in 2026 | Ben Vinegar (Co-founder, Modem.dev) Said 30 Jul 2026
Programming languages that LLMs do not choose, like Ruby and Zig, look dead in the long term.
it does seem like long-term Ruby and Zig are dead. And what I mean is that if the LLMs are not choosing them over a long enough timeline, they're going to become like Louisiana French.
State of Agentic Coding #7 with Armin and Ben Said 12 Jun 2026
Agents fail on niche runtime problems partly because nobody writes up those problems on Stack Overflow or in blog posts any more.
It was a Cloudflare worker oddity. And you know what? There's no more Stack Overflow. There's no more people writing blog posts. There's no one writing, gee, I hit this painful problem with Cloudflare workers and here's how I solved it. So how does the agent index this?
State of Agentic Coding #7 with Armin and Ben Said 12 Jun 2026
Good primitives and libraries unlock more of what is possible than AI coding does; nobody needs another agentic coding multiplexer.
Well, look, we don't need 400 3-column agentic coding multiplexers. We don't need that. The primitives are the bigger unlock to what's possible than the AI coding.
State of Agentic Coding #7 with Armin and Ben Said 12 Jun 2026
Software is going through a fast-fashion phase where projects are thrown away if nobody uses them at once; builders should have conviction and work on things longer.
this like fast fashion, like fast fashion of software is what we're experiencing right now. where people are building something and they're like, hey, look at this. And if they don't, like, if people don't use it, they throw it away and go and try to find some other flash in the pan thing. Folks, my advice to you is have some conviction and work on things a little bit longer.
State of Agentic Coding #7 with Armin and Ben Said 12 Jun 2026
Running agents in YOLO mode lulls people into a false sense of security; everything an agent does should be gated.
you get lulled into this false sense of security. But I've already experienced agents kind of like reason to themselves about why they gotta go and do something and seen, um, seen some bad effects of that. And I just, like, my position and the position of the company, for example, is we just, everything is gated. We don't just let agents go YOLO.
Draw in your terminal! tmux + Terminal Maxxing w/ Ben Vinegar (Syntax) Said 8 Jun 2026
Coding agents make doing things on Linux far more approachable.
I just wanna make a comment that like coding agents make you doing stuff on Linux way more approachable.
Draw in your terminal! tmux + Terminal Maxxing w/ Ben Vinegar (Syntax) Said 8 Jun 2026
When debugging with AI it is easy to get stuck inside the model's search space, chasing its candidate fixes while the real answer sits in context only you have.
A pattern I'm seeing with AI debugging: it's easy to get stuck inside the model's search space. So you burn tokens & time chasing candidate fixes, while the real answer sits in context only you have ... but never explored, because you quietly surrendered your thinking.
@benv.ca on Bluesky Said 2 Jun 2026
AI sometimes makes things too easy, and there is value in the pain.
AI sometimes makes things too easy, and there's value in the pain.
State of Agentic Coding #6 with Armin and Ben Said 11 May 2026
LLMs love code with explicit inputs and outputs and no side effects, so the functional programmers were right.
Hate to say it, but the functional programmers were right. LLMs love code with explicit inputs/outputs and no side effects.
@bentlegen on X Said 15 Dec 2025
Demand for open source feels lower than it has in a while, because people can generate the code they want and companies will use whatever gets them to market fastest, black box or not.
Feels like the demand for OSS hasn't felt this low in a while: 1. People can just codegen what they want 2. Companies will use whatever gets them to market faster to capitalize on AI – if it's a black box so what
@benv.ca on Bluesky Said 18 Nov 2025
AI will write the majority of code.
I've come to accept AI will write the majority of code.
@benv.ca on Bluesky Said 30 Sept 2025
The developer's job increasingly becomes translating what you are thinking into a form agents and LLMs understand; tools will improve, but not to the point of reading your mind.
IMO the job increasingly becomes translating what you're thinking to a form the agents/LLMs understand. The tools will improve, sure, but I wouldn't bet on a future where they know how to read your mind.
@benv.ca on Bluesky Said 23 Jul 2025
When a growing SaaS product's architecture breaks down, the developers he has seen succeed are the ones who fix it one piece at a time, not the ones who want it rebuilt.
When SaaS products grow and the architecture starts to break down, two types of devs emerge: 1. Those who believe the system is beyond repair and needs to be totally rebuilt 2. Those who put their heads down and fix it one piece at a time I've only seen group #2 be successful.
@benv.ca on Bluesky Said 20 Dec 2024
API stability is an underrated killer feature of frameworks and libraries.
"API stability" is an underrated killer feature of frameworks & libraries, IMO.
@benv.ca on Bluesky Said 19 Dec 2024
React's single greatest accomplishment was convincing web developers to stop trying to separate business and presentation logic on the client.
I think React's single greatest accomplishment is that it convinced web developers to finally stop bending over backwards in trying to separate business and presentation logic on the client.
@benv.ca on Bluesky Said 23 May 2023
What is a korrent?
A korrent is a belief a person has stated in their own words: one sentence stating the claim, backed by a quote and a source, kept at korrents.com.
Under a name here, the quoted block is what they actually said. The korrent beneath it is the claim those words support, in korrents' wording — tap it to see the record, its source, and who else holds it.
Nobody here wrote their own korrents. They are compiled from public statements, and a person can change their mind, which is recorded too.
Where they publish
Site benv.ca Their own site: who they are, what they wrote, and where else they publish.
The front door they maintain themselves. It links every account listed here, the books, and the interviews they sat for, which is what makes the rest of this page checkable.
Link verified 7 Sept 2026.
Blog Syntactic Vinegar Six essays on their own site, the most recent from May 2025.
Long, unhurried essays on software careers, self-hosting and startup scenes. There is no feed, so the six posts are listed here as a fixed shelf rather than updating on their own; the newest is from May 2025.
Links
- They expect us to keep changing 14 May 2025
- Now or never now 8 Nov 2024
- Counterscale and the new self-hosted 24 Jun 2024
- The hardest problem 4 Mar 2024
- The community startup scenes we lost 12 Dec 2023
- The story of Third-party JavaScript 3 Oct 2023
Link verified 7 Sept 2026.
x.com @bentlegen Where they think out loud about agents, software and their own trade.
Short posts, often about what building with coding agents is actually like, and the trade-offs they see in it. The most current of their own-words channels.
Recent
- wow it's beautiful, I love it incredible that Apple was producing this kind of aesthetically pleasing hardware in 1984 22 Oct 2026
- it brings me tremendous peace that not only did Sentry become a huge software company but it did so without abandoning its values or forgetting where it came from a real unicorn company 21 Oct 2026
- this is what @modemdev achieves for companies - rapid product iteration if you feel your team isn't there, you should talk to us 21 Oct 2026
Show 17 more
- Reverse-draining Kahwi is a Canadian heritage moment 21 Oct 2026
- Walking home I saw a man smoking meth I wondered if it was because the province had closed a safe consumption site nearby So I had Claude map all the sites Sure enough, one used to be right across the street 20 Oct 2026
- i get lots of cold emails with the same pretend-to-know-what-you-do AI intros do these work? 20 Oct 2026
- fun fact big SF tech corps buy up daycare slots like seasons tickets for pro sports teams so if you have a kid and are looking for daycare, you'll quickly learn all the spots are already spoken for this is part of why people leave 20 Oct 2026
- SF greatly overestimates how much free time your avg person has to use AI People work all day, they commute, they have families and chores and responsibilities ... and at the end of the day many are mentally and physically exhausted And they're not gonna unwind by prompting 20 Oct 2026
- hunk v0.24.0 is out - with improvements from 20+ contributors 🔥 including: - open any file or directory with `hunk open` - merge conflicts now render correctly - configurable auto layout - CLI tab autocomplete, scroll_off, and much more 19 Oct 2026
- my name is Ben and I graduated from one of Canada's top computer schools with really good grades now I help software businesses find problems buried in their customer channels 19 Oct 2026
- Went to a bookstore this weekend and brought home this beautiful time capsule of Toronto 19 Oct 2026
- whoops I asked the agents to emulate a real human engineering team so they clocked out until Monday morning 😅 18 Oct 2026
- new AI software factory experiment, modeled after a human eng team: - multiple pi-durable agents - each w/ their own desktop linux VM (8gb ram) - each has a role/persona, works towards common goal - private IRC server for collab - eng rituals: daily standups, "review my pr", etc 18 Oct 2026
- I like posting and reading content on X But if you rely on this platform to tell you the truth about the world, you will become increasingly divorced from reality 18 Oct 2026
- if you invested $70k in the Nasdaq in 2017, it would be worth $450k USD today 18 Oct 2026
- last modem I ordered off eBay got lost in the mail - soured me on getting another for a while BUT we had a really good week so this guy is coming home 🤞 16 Oct 2026
- it's unrealistic for engineers to monitor every support channel - it's noisy, it's distracting, they have their own work to focus on but the best engineers want to fix user problems! so instead just bring them the important stuff, automatically 16 Oct 2026
- Don’t read the code Don’t look at the docs Don’t watch the explainer vids Don’t learn the UI Don’t even use the software 15 Oct 2026
- ChatGPT yearns for Yellowknife, NT 15 Oct 2026
- when I look at the biggest mistakes we've made over the past year right up there is "rolled our own software" even if you can make a decent clone, the time maintenance is so onerous - I wish I had just swiped a credit card instead 14 Oct 2026
Link verified 7 Sept 2026. Recent items update automatically from the channel.
Bluesky benv.ca The same short posts, on Bluesky.
A Bluesky account whose handle is their own domain, carrying much the same short posts as the X account.
Recent
- 🆕 oh yeah hunk v0.24.0 came out over 50 fixes from 22 contributors, plus: - open any file or directory using `hunk open` - merge conflict support - auto layout is now configurable - terminal tab-complete and more full notes at 21 Oct 2026 hunk.dev
- 🖥️ new project: Glance Desktop® OSS native screenshot tool built in Rust + GPUI it's like regular screenshot tools, but with: - one-click links to share imgs w/ remote agents - fancy GPU-rendered looping bgs, export to mp4/gif ⬇️ 4 Oct 2026
- 🎨 released hunk v0.23.0 yesterday: - proper text search via `/` - instant GitHub PR diffs via `hunk gh pr` - your terminal colours right in hunk - open files in Zed - plus fixes & perf improvements 1 Oct 2026
Show 17 more
- Opus 5.5 product vids aren't just good because they look great they're good because they can explain the product better than you can 28 Sept 2026
- I recently learned there are way more monospace fonts out there than I was aware of somehow this led to making a video that tours 18 different fonts in 3 mins, set to chill lofi music (🔈) hope it's useful for folks - it was for me! 22 Sept 2026 youtube.com
- 🔧 tune your AGENTS.md in 5 mins go to ossrules.md, copy the prompt, hand it to your agent your agent will browse our index of OSS project rules and see which apply best to your repo 22 Sept 2026
- 📜 introducing ossrules.md - learn how leading OSS projects use agent rule files - discover skill files that projects actually use - level up your own projects and workflows I've personally learned a lot just exploring 18 Sept 2026
- today an engineer found a bug where we over-billed some event types, and made a fix Modem pinged him - a customer flagged this 2 weeks ago, but it got lost in the noise he had no idea! so he followed-up & let them know it was addressed 5 mins of setup → never miss a follow-up 16 Sept 2026
- 🏛️ announcing hunk v0.22.0 - now w/ `hunk log` - TUI for reviewing git/jj history - quickly jump from commit → diff - OR multi-select a commit range - supports mouse + all colour themes also in this release: multi-line comments, lower memory usage, more efficient --watch mode, and more! (🔈) 10 Sept 2026
- the only cli benchmark that matters 7 Sept 2026
- 👀 here's another take on browsing git history inside hunk: hunk-history (extension) ⬇️ looks pretty nice 6 Sept 2026
- we're now processing a lot of data, which is great but the days of infrastructure easy mode have come to an end 3 Sept 2026
- 🎙️ State of Agentic Coding w/ @mitsuhiko.at and @benv.ca returns In this episode (#9): - how agents are getting more autonomous - the subsidies will continue until morale improves - herdr and VC-funded terminal apps - vibe coding from your car, and more! Available on YT, Spotify, and via RSS ⬇️ 17 Aug 2026
- It's becoming normal practice to plug your coding agent into tools like Slack and helpdesk software to find customer insights. Turns out it's super inefficient, because you're just fetching and smashing raw strings into your context repeatedly. So we released a new MCP server that solves that. 31 Jul 2026
- ✅ Yes the code matters In experiments we ran, well-named and organized code can dramatically improve agent outcomes: - fewer tokens spent/more context window preserved - fewer confidently wrong answers developed Verified against 14 model/harness combos incl Sol, Fable, GLM 5.2 and more 24 Jul 2026
- 👋 New blog post: "How coding agents read files – and how to write for them" In it I explain how code optimized for agent-discovery can lead to major token savings on search + retrieval, even with frontier models like GPT-5.6 Sol Plus other important findings ⬇️ 20 Jul 2026
- 🎙️ State of Agentic Coding #8 with guest Mario Zechner is out. In this episode: - Vibe-checking Fable & GLM 5.2 - @mitsuhiko.at teaches us RL - How harnesses smooth out model jank - Loops, surviving AI fomo, and more YT: www.youtube.com/watch?v=_lfp... Spotify: 14 Jul 2026 open.spotify.com
- 🛠️ hunk 0.17.0 is out Patches from 10 new contributors, w/ small fixes & quality-of-life improvements like: - official homebrew install (brew install hunk) - launch hunk w/o menu bar - better session reliability - esc no longer quits, proper transparent bg colours, & more 7 Jul 2026
- Just noticed a vibe-coded project of mine cost $200/month to serve a few dozen users/day 😭 Turns out it was doing 10s of millions of unnecessary Redis polls/day, even when idle - telemetry alone cost $$ a possible clue for why infra providers are growing so much right now 28 Jun 2026
- there's a lot of tech companies in Toronto now 13 Jun 2026
Link verified 7 Sept 2026. Recent items update automatically from the channel.
Podcast State of Agentic Coding Co-hosted monthly with Armin Ronacher. Two hosts talking, so it is their own voice.
A monthly conversation with Armin Ronacher about what coding agents can and cannot do yet. A co-hosted show rather than an interview programme, so both voices are the point; nine episodes, the newest from August 2026.
Recent
- State of Agentic Coding #11 16 Oct 2026
- State of Agentic Coding #10 9 Sept 2026
- State of Agentic Coding #9 17 Aug 2026
Show 6 more
- State of Agentic Coding #8 13 Jul 2026
- State of Agentic Coding #7 12 Jun 2026
- State of Agentic Coding #6 11 May 2026
- State of Agentic Coding #5 10 Apr 2026
- State of Agentic Coding #4 12 Mar 2026
- State of Agentic Coding #3 16 Feb 2026
Link verified 7 Sept 2026. Recent items update automatically from the channel.
Interviews As a guest Interviews they sat for on other people's shows -- chosen, not exhaustive.
Their own voice by their own participation: long conversations where they were the one answering, on the interviewers' own shows. Listed on benv.ca, and a curated few rather than a complete file.
The conversations
-
Ben Vinegar × Distributed Tracing and TypeScript Migrations
Syntax #434
23 Feb 2022 -
Ben Vinegar, Sentry – reflecting on ladders, acquisitions and trajectory
Engineering Unblocked
29 Sept 2023 - The New Definition of Frontend Development Code[ish] 85
Link verified 7 Sept 2026.
-
Ben Vinegar × Distributed Tracing and TypeScript Migrations
Syntax #434
GitHub benvinegar Their code, including Counterscale, the analytics tool they wrote about.
Forty-five public repositories, among them Counterscale, the self-hosted analytics project one of their essays is about.
Recent
- pi-whimsical-toronto — Toronto-inspired working messages for the Pi coding agent 6 Oct 2026
- omarchy-couch 6 Oct 2026
- termdraw — Agent-friendly ASCII illustrator for the terminal 4 Oct 2026
Show 12 more
- counterscale — Scalable web analytics you run yourself on Cloudflare 4 Oct 2026
- prompt-cardio — Prompt Faster: the vibe coding interview. A typing test disguised as an AI chat. 13 Jul 2026
- opentui-island — Embed your OpenTUI components in Ink and pi-tui 19 Apr 2026
- ansi-to-img 18 Apr 2026
- dotfiles 16 Apr 2026
- piday 16 Mar 2026
- pi-stuff 15 Mar 2026
- piductor — Terminal-native conductor for parallel Pi workspaces 10 Mar 2026
- txjs-1997 — This is the source code for the 1997 Texas JavaScript conference home page. 10 Mar 2026
- chrome-ai-post-detector — Chrome Plugin for identifying AI-authored posts on X 5 Jan 2026
- nitrotype-tps — Vibe coded Chrome plugin that converts WPM in Nitrotype to TPS (tokens per second) 23 Nov 2025
- useful-scripts 11 Aug 2025
Link verified 7 Sept 2026. Recent items update automatically from the channel.
Books Books Third-party JavaScript, and a chapter in Beautiful JavaScript.
Third-party JavaScript (Manning), written with Anton Kovalyov, on building code that runs inside other people's pages; and a contributed chapter in O'Reilly's Beautiful JavaScript.
The books
- Third-party JavaScript Manning
- Beautiful JavaScript O'Reilly, contributor
Link verified 7 Sept 2026.
About the English under a post
Some people here publish in a language other than English. Where they do, this site shows a machine translation beneath the post, in this typeface — the site's own, not theirs.
The post itself is never changed, moved or hidden: what is set in the serif above is exactly what the person published, and it is what to quote them on. A translation can be wrong in ways that matter, especially about tone.
Only the post's own words are translated. A quoted post, a linked article and a belief on korrents.com are left in their original language.