ben 🐝 human 🧬 AudioTrack 🎺 AI Tinkerers Seattle Chapter 🤖 OpenAI Ambassador 🧑💻
Seattle
Joined October 2011
- Tweets5.7K
- Following778
- Followers985
- Likes43.6K
Ben Steinher 🐝 retweeted
It's time to kill the API key.
Last week, a hacker spent $600 on an API key leaked by the host of my friend's vibecoded app.
A lot of people learned last week how dangerous it is to "litter" their API keys all over the place.
The truth is, API keys aren't the right fit for vibecoders building personal apps. They weren't even the right fit for trad developers.
API keys aren't just insecure by default, they're also annoying to use. We've only put up with them this long because, pre-AI, developers spent way more time coding than fiddling with API keys.
Today, when you @usecharming to build and publish apps, you can use +1500 connectors without api keys. Our goal is to make that 100%.
Connectors are more secure by default, and we are also working closely with the @composio team to make them even safer than API keys.
Ben Steinher 🐝 retweeted
Do mentors without $300 an hour Calendly links exist?
Looking for mentors in the social/digital/brand marketing spaces. If in tech is a plus.
Criteria:
Must be a leader in the space, have worked or work with big brands
Must be 5-10 years my senior. I have 10 years experience.
Ben Steinher 🐝 retweeted
Agents can build a whole app's UX now. What if stakeholders could review all the UX changes at once?
To go from a V1 to a great app, you need to iterate until it's amazing. Use Decode, work together, and get there the fastest.
Ben Steinher 🐝 retweeted
when nick told me that he had some interp breakthroughs, he was uh, underselling it. i think this may be remembered as a major moment in scientific history. news so good even eliezer will have to lower his pdoom. we can start making alignment into a 'normal science'
interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had
with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works
already-clean.goodfire.com/
Ben Steinher 🐝 retweeted
My biggest obsession right now is doing work with my AirPods in: walking and talking to a "coordinator" agent who can do tasks in parallel.
When I'm sitting at a desk, my left brain is activated.
I'm in logic and facts and procedure mode.
When I'm walking, I activate more of the right brain. Something about moving the body lifts my spirit. I'm not only more creative... I also have more energy. My mood is lifted, and I can think more clearly. I even write better while walking.
This has been a habit of mine for 2 years now. I would take voice memos while on walks, and then, when I returned to my desk, I would have a "plan." Well, now I can literally do things while walking. My agent (Dot, currently the best real-time voice agent) can spin up local Codex sessions on my computer, is connected to all my tools, and can communicate with anyone on my team. All by simply talking.
(sent via Riley's Dot using the x-api skill)
Ben Steinher 🐝 retweeted
Six demos at @AITinkerers Seattle Games & AI night. The number that stuck: studios spent $25B on player acquisition in 2025 for 0.8% install growth. The answer on stage was the real game, playable in a browser tab. 60-second recap.
Ben Steinher 🐝 retweeted
Meet our Codex Ambassadors 💙
We sat down with a few of our ambassadors from around the world to hear why they joined, what they’re building, and what makes their local communities special. Proud to support them!
Join our Codex Ambassador program: developers.openai.com/commun…
Ben Steinher 🐝 retweeted
the best software is going to be the kind we make for ourselves and other people just happen to love too.
Ben Steinher 🐝 retweeted
Codex Ambassador applications are now open! Come join our global ambassador community 💙
Already showing people what they can build with Codex? Applications to become a Codex Ambassador are open again.
We support you in hosting meetups or sharing what you’ve learned:
developers.openai.com/commun…
Ben Steinher 🐝 retweeted
we created Codex Ambassadors in february with 15–20 people in mind.
now we have 199 ambassadors across 49+ countries & territories who’ve run 300+ events (!!)
waking up to photos of people building together in Tokyo, India, London, SF has been one of my favourite parts of building this programme.
really proud of the community we’ve built.
applications are open again, come join us 🙌
openai.com/form/codex-ambass…
Already showing people what they can build with Codex? Applications to become a Codex Ambassador are open again.
We support you in hosting meetups or sharing what you’ve learned:
developers.openai.com/commun…
amazing program, really enjoy waking people up to all the things you can do with ChatGPT and Codex.
last meetup I used voice mode to translate a 15 minute conversation with someone who spoke spanish, so his son could go enjoy the demos, and now with zero prior experience he’s already contracting multiple websites for old companies he used to work for.
anyways, it’s really great to see the ah hah moment happen in realtime.
if you like to shepard people into the future, you should definitely apply, we’d love to empower your local community.
Already showing people what they can build with Codex? Applications to become a Codex Ambassador are open again.
We support you in hosting meetups or sharing what you’ve learned:
developers.openai.com/commun…
Ben Steinher 🐝 retweeted
I gave my dad access to Codex, and he’s building himself an options forecasting app.
The era of personal software is here!
i’ve been loving the proactive unlocks this grants my dot
MCP Events lets you build automations that trigger work in ChatGPT when something happens in your app.
Think of workflows like:
- When a bug report lands, investigate it and prepare a draft PR with tests
- When someone comments on this proposal, work through the edits and show me what changed
- When a new podcast episode drops, pull out the ideas relevant to my project
Essentially you're giving ChatGPT instructions upfront. A matching event can start the work, without you having to notice the update, copy it into a chat, and prompt again.
Give this a try with a plugin that supports events!
Ben Steinher 🐝 retweeted
One week since we launched dots!
We’ve been working through your feedback and shipped a bunch of improvements:
- Faster cloud browsing
- Fixed extra phone notifications while you’re already using ChatGPT
- Cleaner code blocks and attachments on web
- Fewer pauses while waiting for a reply
- Faster sidebar loading, smoother animations and less loading jank
- Fixed character rendering in Safari and Firefox
- Fixed Outlook setup getting stuck
- Clearer plan guidance + a way to retry failed activity loads
- More stable safety banners on web
- Fixed clipped approval buttons and oversized Gmail previews hiding approvals
- Fixed enterprise access issues on web and when starting voice
- Better dot organization of when to use new versus existing codex threads
Up next:
- Voice reliability (being improved everyday, but we’ll shout when it’s up to par with our expectations)
- Increasing dot’s control of your Codex app: manage project folders, have full Codex thread context, rename threads, and more
- Better organization to know what your dot is working on at any given moment
- Connect your dot to multiple computers
- Better Windows support for local computer control
- what else?
Thanks for trying our product and giving us lots of feedback. Keep it coming!
Ben Steinher 🐝 retweeted
I'm so sick and tired of my Sonos speakers never working correctly (the last straw today was trying to get Muse to act as a DJ for me using my Spotify likes as a guide) that I decided to just say screw it and started FrankenSonos.
The goal is to just reverse engineer everything and do it correctly using Rust and all my libraries like asupersync and so forth.
I'd also like to get it to all work seamlessly with Tailscale if I can. Maybe I'll be able to make my own FrankenFirmware for them and then I can finally be rid of their garbage software completely.
I have a mixture of older S1 hardware and the newer S2 hardware in my house, so I'm being careful to support all of it seamlessly in a unified, coherent way.
The irony is that they are incredibly awesome-sounding speakers when they actually DO work. And you can buy them on eBay pretty inexpensively. The software is just so bad and broken that it ruins everything.
You can follow along here:
github.com/Dicklesworthstone…
