Evil Martians (@evilmartians) header
Evil Martians (@evilmartians) profile picture

Design and engineering consultancy for devtools, AI, and cybersecurity startups. Lago, Tines, Teleport, and Whop grow with Evil Martians. We host @sfrubyconf.

New York, NY
Joined August 2008
Photo by Evil Martians (@evilmartians)Photo by Evil Martians (@evilmartians)Photo by Evil Martians (@evilmartians)Photo by Evil Martians (@evilmartians)Photo by Evil Martians (@evilmartians)Photo by Evil Martians (@evilmartians)Photo by Evil Martians (@evilmartians)Photo by Evil Martians (@evilmartians)Photo by Evil Martians (@evilmartians)
Rails + AI models, which to use? Claude Opus 4.5 is slightly more accurate and less expensive than Fable, although a tad slower. However, you can get essentially the same results with Kimi K3 for almost half the price. We know this because we worked with @Rails to benchmark the performance of eight AI models against 21 atomic tasks and evals. This project gave us the amazing opportunity to work hand in hand with @AmandaBPerino and the Rails Foundation on the creative task of building Rails-centric LLM benchmarking. But also, this helps move the ecosystem forward and hopefully allows the community to keep ideating with Rails. Congrats to @skryukov_dev and Artur Petrov for their work on this!
9
7
1
98
8,003
Evil Martians retweeted
Four new versions have been benchmarked in Agents on Rails: @OpenAI's GPT-6.2 Astra (xhigh and medium) and GPT-6.2 Sol (xhigh), and @deepseek_ai V4.1 Flash 1017 (max). Here’s what we found: New first place: GPT-6.2 Astra. At xhigh effort it scores 65.8%, narrowly ahead of Claude Opus 5.5 (65.3%), at $4.42 per run versus $10.53. It’s also faster than the previous Astra. At medium effort it lands at 4th place and averages under 11 minutes per run, the fastest configuration we’ve tested. GPT-6.2 Sol is nearly twice as fast as GPT-6.1 Sol (about 12½ minutes per run versus 23½) at roughly the same cost. Its accuracy dipped this round, from 60.8% to 54.1%, though some of that may be run-to-run variation. DeepSeek V4.1 Flash 1017 is a big efficiency gain. It scores within a few points of the previous version while running about 1.5× faster at roughly one-sixth the cost. At $0.16 per run, it’s the cheapest model we’ve tested to date, but you get more accuracy and speed with Luna or Sol for just a little more cost. Thanks once again to the team at @evilmartians for their work. Explore the full results at: rubyonrails.org/ai
8
8
3
52
4,570
AI agents work best with opinionated instructions, and Martians have plenty  Teach your agent to build apps the Martian way by using our 13 skills based on real project experience: design systems, layered Rails, Inertia, Tailwind, anti-slop guardrails, LLM visibility, and more. These all work with Claude Code, Cursor, Codex, Copilot, and Gemini CLI: evilmartians.com/agent-skill…
12
6
23
878
Evil Martians retweeted
Hey everyone, Ruby 4.0.10 is out! It contains a critical security fix. Update immediately! ruby-lang.org/en/news/2026/1…
12
178
11
503
55,873
Evil Martians retweeted
Coming full circle: Last June, Thicket had its very first seminar. Greg Salyer (PhD Emory) led us through cognitive anthropology, religion, mythology and more. The course started with a simple insight: “Humans Anthropomorphize” Last week, Travis Turner of @evilmartians (our friends who helped build Thicket) gave amazing talk at @sfrubyconf and took this insight in a very different direction.
1
6
1
12
509
"Got AI Ontological Shock?" is the short title of the talk our editor-in-chief, Travis Turner, gave at @rmrubyconf a few months ago. The thesis is that the uncertainty, fear, or discomfort many of us feel around AI is "ontological shock": our brains can't fully process what's happening. To @travis____t the answer is community and having open conversations about it. This talk has gotten a lot of praise, and we highly recommend watching it: youtube.com/watch?v=nzg03qHC…
2
8
544
This abduction was consensual!
Got abducted by @evilmartians Joining the team behind PostCSS, nanoid and nanostores as a frontend engineer. Let's go!
3
23
993
We have community. Last week, hundreds of Ruby on Rails engineers met in San Francisco to chat about the future of the ecosystem. The @sfrubyconf was even better than what we envisioned, and we’re beyond grateful to everyone who played a part in making this happen. 💖 Like @garrytan said in his opening words: “The best years of Ruby are still ahead of us.” Let’s keep building together. See you next year: luma.com/sfrubyconf2027
9
2
36
2,062
A little BTS for @sfrubyconf 2026 featuring a bunch of @evilmartians and our speakers, the Pier community puzzle and the beautiful SFJAZZ venue ✨
7
24
1,067
When coding agents start making design decisions, it's common for UI drift to appear. That’s when our team comes in to clean it up. We take a thorough inventory, build a design system that both engineers and agents can use, and add guardrails to keep drift from coming back. @buildkite and @currents_dev have already worked with us on this. Check out this service and book a call: evilmartians.com/services/ai…
1
6
18
1,058
Buildkite is a CI/CD platform used by engineering teams at Canva, Pinterest, Slack, and PlanetScale to automate building, testing, and deploying software. We're currently building a unified design system and UI kit across all its products: evilmartians.com/clients/bui…
1
1
6
328
Currents is a test observability platform for running, debugging, and analyzing Cypress and Playwright suites in CI. Its engineers had been making design decisions as they built and grew, which ended up introducing UI drift. Our team audited the platform, redesigned key screens, and built a design system that engineers and agents can ship from: evilmartians.com/chronicles/…
2
243
Evil Martians retweeted
Shout-out Saturday, @sfrubyconf edition. The people who made my week better: @garyhtou for leading the scholars program and inviting me to join. @AmandaK14007213, @inazarova and the entire @evilmartians team for a first-class conference. @vmelnikova_en for the insightful VC pitch session. @nicoerlichman for the fantastic emceeing. @palkan_tula for keeping the mood light during #SFORubyConf And all the other new connections I made at SF Ruby. Such a great time. Thank you.🙏 
6
4
30
921
Evil Martians retweeted
Incredible conference. Incredible people. Incredible time. Incredible!! Thank you @sfrubyconf! Can’t wait for next year!! As always, it was such a pleasure to see a tiny snippet of how your brain works @inazarova thank you so much for everything It was sooo good to finally meet @vmelnikova_en @CamiMirabalMktg hope to see you again soon
4
2
27
805
We have been building and growing Rails applications since 2006. We've done it all: MVPs, APIs, billing systems, admin dashboards, upgrades, migrations, major fixes, you name it. Our team comes in when you need us the most. We solve your problems, train your people, and help you hire if needed, so your team keeps succeeding even after we end part ways. A message from our of our former clients, @Synadia, who hired us for Rails consulting:
3
3
15
795
Evil Martians retweeted
5 tips for designing fair, reliable black-box tests, based on what we learned from working on the Agents on Rails benchmark project with the @Rails Foundation. But first, some context: Agents on Rails measures how well AI models handle real Rails coding tasks. It happened in stages: - Stage 1: Evaluating models on atomic API-level tasks - Stage 2: Testing how well models handle single feature tickets - Stage 3: Asking models to complete multi-step tasks to build or maintain a product Read this if you’re planning to build your own LLM benchmark: 🧵
1
6
666
Evil Martians retweeted
Chris Oliver speaking the truth @sfrubyconf closing keynote. @excid3
5
31
1,173
Evil Martians retweeted
Last year the Rails-powered creator platform @uscreentv shocked us by raised a $150M investment. Then, Uscreen partnered with @evilmartians to build a new headless product to support the most AI-forward of the customers. What we built and how - read on the new blog post: evilmartians.com/chronicles/…
3
1
17
1,112