@thsottiaux ---
22/22
My advice: Focus. Simplify. Build. Deliver. OpenAI doesn't need to win every argument online. It needs to build products so good that the arguments become irrelevant.
@thsottiaux 1/22
OpenAI sometimes looks like a company trying to do everything at once: too many directions, launches and reactions, while its most important products still aren't getting the focus they need to become exceptional.
@thsottiaux ---
21/22
Listen to users and fix real problems. Someone who has used ChatGPT for seven months and dislikes a change deserves to be heard. But one person's reaction shouldn't dictate long-term strategy. Separate genuine feedback from social-media noise.
@thsottiaux ---
20/22
One of OpenAI's biggest challenges is its public narrative. Too often, the conversation is shaped by outrage, oversimplification and personality-driven controversy rather than the technology and work itself.
@thsottiaux ---
19/22
I'm not suggesting OpenAI should sacrifice important capabilities to win social media. I'm saying product progress, public perception and market adoption are connected. Understand that connection, but don't let it dictate the roadmap.
@thsottiaux ---
18/22
That's how attention turns into adoption. A 12-year-old sees the coding buzz, tells their family ChatGPT is the subscription to buy, and influences the decision. Public conversation shapes what people notice, compare and pay for.
@thsottiaux ---
17/22
The reverse can happen: coding improves substantially, while writing or video gets worse. If coding is what people are watching, the reaction may still be positive - even though the model has regressed in other important areas.
@thsottiaux ---
16/22
Consider this: a model gets much better at writing books but worse at coding. The writing improvement is real and valuable. But if coding dominates the conversation, people may focus on the regression and call the release a loss.
@thsottiaux ---
15/22
Coding is one of the most visible battlegrounds in today's AI conversation. It shouldn't be the only priority, but OpenAI can't ignore how strongly it influences comparisons, reputation and buying decisions.
@thsottiaux ---
14/22
Above all, build a model that's affordable, fast, reliable and genuinely capable. Then pay close attention to where people compare AI products, what they discuss, and which capabilities shape their decisions.
@thsottiaux ---
13/22
Reassert OpenAI's position as an AI industry leader through execution. Build constructive relationships with governments, earn institutional support and demonstrate long-term value. Don't turn every challenge into a public spectacle.
@thsottiaux ---
12/22
Make the products people use most exceptionally reliable: ChatGPT, coding tools and the workflows built around them. Prioritize speed, stability, consistency and safety. Improve what people already depend on before adding more.
@thsottiaux ---
11/22
Drop unnecessary drama and performative communication. Be direct, calm and confident. Less gimmickry, less trying to be everyone's friend. A company at the frontier should let its work earn respect.
@thsottiaux ---
10/22
Stop teasing. Start delivering. Announce less, build more, release when something is ready, and let the finished product speak for itself.
@thsottiaux ---
9/22
Stop prioritizing features with little demonstrated demand while core products still need work. Innovation matters, but shipping more things is not the same as making the product better.
@thsottiaux ---
8/22
Avoid unnecessary resets and reversals. Users need predictable limits. Constant changes may soothe today's complaints, but erode trust. A sustainable product won't satisfy everyone; some customers may leave. Better that than teaching users the rules never last.
@thsottiaux ---
7/22
Go quiet publicly, not internally. Keep listening, study the criticism, fix genuine problems and focus on the product. Come back when there's something substantial to show, not just something new to announce.
@thsottiaux ---
6/22
Pause OpenAI's social-media posting for three months. Stop trying to join every argument or answer every provocation. Right now, public engagement seems to be generating more hostility than goodwill.
@thsottiaux ---
5/22
Rethink usage allocation. Keep chat and coding under a clear, coherent system, but give users much more coding capacity - potentially 3x the Codex-equivalent allowance. People should be able to do substantial work without constantly hitting a wall.
@thsottiaux ---
4/22
Set sensible limits on image generation if doing so frees resources for the core experience. Expensive capabilities shouldn't crowd out investment in the model and the products people rely on every day.
@thsottiaux ---
3/22
Dots: reconsider making it a public product. If it's genuinely valuable, run it internally at high speed on your best model and use it to improve how OpenAI operates. Not every powerful internal tool needs to become a public product.