research, product and finance @openai
San Francisco, CA
Joined July 2026
- Tweets492
- Following100
- Followers4.9K
- Likes1.7K
Pinned Tweet
As promised, my 𝗖𝗵𝗮𝘁𝗚𝗣𝗧 𝗪𝗼𝗿𝗸 onboarding guide.
A companion to the tips so far, with more to come!
Use it to build a setup around your actual job, whether you’re starting fresh or improving how you already work
work-guide.openai.chatgpt.si…
“Helping everyone in the world have equal access and opportunity”
Odd to say this was her work motto early in her career, then take such a hard stance against ads later in the interview
Ads help fund free access to ChatGPT at global scale. Not everyone can afford a subscription
Dario reveals he was never interested in starting a business before Anthropic
"I don't think either of us imagined that we would start a company. That's not really how either of us thought of our lives or about what we wanted to do. I had always wanted to be a scientist. I actually found companies in business to be of very little interest to me."
Daniella Amodei: “I would say Dario and I were very unusual in that, from a really early age, we just genuinely really liked each other. We were always just best friends since we were really little.”
“I don't think starting a company was ever on our bingo card. Neither of us were particularly company oriented."
"I would call myself a classic do-gooder. I worked in international development. I felt really strongly about helping people everywhere in the world have equal access and opportunity.”
“And I became a vegetarian when I was quite young. It's the only thing that I did first. Dario did everything first because he's older, but I became a vegetarian and then he became a vegetarian."
Dario: “I think I was three years later or so. It seemed convincing to me."
Kyle Kober retweeted
If you didn’t try Terra because the old AA score didn’t impress you, I think you care about benchmarks too much.
Most evals do not measure things that matter to how you actually use models and many are deeply flawed.
Back in my investment banking days, it was embarrassing to get caught using a mouse. You HAD to know the keyboard shortcuts
Given the trajectory of computer use, maybe in 2027 we’ll just talk to our computers all the time and it’ll be embarrassing to own a mouse or keyboard 😂
I've been using 6.1 Terra all day editing and I can confidently say I will be maining it for editing. It's unbelievably good at knowing exactly what to do when I prompt it for an edit, way more consistent than Astra. It's blazing fast at the lower reasoning levels with no clear difference in intelligence editing-wise.
Consider me won over idk how they did this...
Kyle Kober retweeted
Astra is the smartest model, and best at visual work.
Sol is nearly as good, but much cheaper.
I understand there's a lot, but they each have different price / speed / perf tradeoffs.
Top left is still my favorite corner 🤓
Terra matched Astra on this test at a tenth of the price per token. I’d try Terra first for “which formula broke?” and save Astra for “the math checks out, so why don’t I believe the forecast?”
We’re judging tomorrow’s economy by today’s job titles
Firms will hire people to adapt, evaluate and maintain AI. Walmart already is. I think this becomes routine, alongside jobs we haven’t imagined yet
Take displacement seriously … but leave room for human ambition too 💪🏻
An older GPU can still be the right GPU 🤓
H100s still do image gen well. H200s bring more memory and bandwidth for text generation
Pairing machines to workloads and optimizing inference can keep older chips productive. Cost per output, at the latency you need, is what matters
a16z's Alex Danco and Moses Sternstein on the bear case that GPUs would depreciate fast, what rental rates actually show, and why the data points to scarcity:
Alex: "There is a whole case that people made on both sides of the argument that these things were assets that would depreciate on a schedule. And in the past 18 months, I think what we have learned is that all our assumptions were wrong."
Moses: "GPU rental rates continue to go up. Even as token pricing comes down, you see more and more consumption, and it's causing relative scarcity amongst the GPUs that we have."
"Even for the older GPUs, the A100s, which have been around already for five years, that were supposed to become obsolete... Whether they had 4-year depreciation schedules, 6-year depreciation schedules, they were supposed to be obsolete almost immediately, and that was the original bear case."
"And plainly, they're not... Everybody is priced above what they should be based on their scheduled obsolescence. And that's true of the older models, the A100 and the H100."
@alex_danco @MosesSternstein
𝗪𝗼𝗿𝗸 𝘁𝗶𝗽: What does the policy mean? 📋
Upload an approved policy. Ask Chat for employee questions, proposed answers and supporting clauses. Flag anything the policy doesn’t settle
Review the edge cases with the owners before turning it into an interactive FAQ Site
𝗪𝗼𝗿𝗸 𝘁𝗶𝗽: Explain that chart 📊
Upload product revenue. Ask Chat for a trend view, a product comparison and a growth breakdown. Have it prove it can explain the "so what"
If it can't, feed it more context to maximize its understanding of (1) process and (2) performance
Kyle Kober retweeted
Day 16/
You can now cache for 12h in the API and for logged in users across Codex & Work we keep cache forever and you don’t have any penalty for coming back to an old conversation or resuming after a few hours and are now guaranteed to hit the cache.
Kyle Kober retweeted
Insane.
My dot found an $8,300 mistake in my tax return. The return was prepared by a tax firm that is best in class!
People, if you’re not using ChatGPT for your Finances, you are definitely leaving money on the table.
𝗪𝗼𝗿𝗸 𝘁𝗶𝗽: Same number, different meaning 🔢
Upload both reports. Before explaining the gap, have Chat compare periods, units, definitions and scope. Then calculate the difference on a common basis
Check the mapping. Sometimes the “variance” is two different questions
Was lucky to be Vicky's "pirate" 🏴☠️ or whatever we're calling a Codex-empowered SME partnered up with amazing SWE ...
MUST FOLLOW 👇
Joined @OpenAI over a year ago and it’s been fun working on agents across ChatGPT to Codex.
We just launched multiplayer slides where everyone and their agents can collaborate.
There’s so much for you guys to discover, and we’re working on getting it to more people :)
Send feedback and feature requests! What do you guys want to see?
Kyle Kober retweeted
Yes! Reusable building blocks and a brand style are exactly the direction we're going. We’re working on templates and brand personalization for Sites. Would love to see your chart/layout kit to understand what you’d want to carry over, and see how a skill could maybe solve the immediate need.
𝗪𝗼𝗿𝗸 𝘁𝗶𝗽: What changed in v7? 📄
Upload the old and new project briefs. Ask Chat to compare scope, owners, dates and assumptions. Have it quote the relevant passages and setup a diff to juxtapose files for review
That's right ... knowledge workers can diff now too 🤙🏻
After ~2 decades making PowerPoints, Sites are the next step. Corporate world is ready to evolve
Your own database, automated updates, custom HTML designs and tools, with sharing and collaboration too
Try the Site plugin in ChatGPT. You have no idea where that’ll take you 🤓
Multiplayer slides. You, your coworkers and your agents working on the same deck
The team cooked on this! Try it out and let us know what you think 👇🏻
ChatGPT Slides has new ways to get your deck ready—and get everyone on the same page.
• Work together in real time in ChatGPT: Share decks, edit together, and leave comments. Ask ChatGPT or your dot to turn feedback into updates.
• Edit right on the slide: Click to change text, update charts, and move or resize images and shapes.
To get started, ask ChatGPT to "create ChatGPT slides" from a conversation or a file you've shared. Present in ChatGPT, or export to PowerPoint or Google Slides.
Available on Pro, Business, and Enterprise plans.
chatgpt.com/features/slides/
