The cascade has been building momentum for awhile now. People have realized that @AnthropicAI is no long on their side. They're going after enterprise and fucking everyone else off. If @OpenAI stays the course (supporting devs and the community) and keeps releasing models like 5.6 and better in support of that, they're going to own the market simply because the *community* stays behind them. I'm saying this as someone who adored Anthropic and now can't believe I was as big a fan of them as I once was. Really sad to see what they've become.
Same boat, largely. Mostly relying on engineering and process knowledge at this point. Your piece from a little bit ago on intent engineering largely gave voice to how I view my role at this point. I know what I want. I know what makes it good, secure, and solid. I just convey that to the agents doing the work and guide them toward the intent. Most of the decisions on how we there are theirs.
I really wish more people were talking about the fact that HuggingFace had to turn to a Chinese model to deal with the breach not because of cost, but because of capability......
Hahahahaha - this is amazing.
Hollywood: AI is dangerous and is going to kill our industry. Our stars need to start trademarking their voices and likenesses to protect our 7- and 8- figure actors.
Also Hollywood: https://t.co/5L9duUxWwU
One of the most interesting realizations you can have here on X is when you realize that most of the AI influencers all follow the same playbook:
1. Hype a thing
2. Wait for the next thing to be released
3. Hype the new thing
4. Wait for the next thing
5. Hype the next thing
6. Repeat ad nauseam
Don't believe me? This month's movements between ChatGPT, Claude, GLM 5.2, and others make for a great opportunity to see it in action.
Pick an influencer account and scroll back about a month or so. Watch as they tell you that <company> just made it game over and this new model is the most insane thing they've ever seen. Then scroll up to the next shift/release. Apparently it wasn't "game over" because <other company> just made it game over and *their* model is insane.
Model releases are cool. AI is cool. But the hype machine.....it's just getting old, boring, and transparent.
@Teknium I don't know you but I appreciate you, your work, and your presence here on X. You're in my thoughts as you go through this difficult time. That loss is devastating (been there). Take time, heal, and know that your dog knew how much he was loved.
Can't agree with this more. While I still use Claude for a few things, I'd estimate that 90% of my AI tasking is run through ChatGPT at this point - not only because of Sol and its quality but because @OpenAI's engagement with the community and lack of playing games has brought me onto their side.
If history is any kind of teacher, that may change. But I really, really hope it doesn't. The ball is literally in their court but there's a real path here toward domination, and it's their treatment of the dev community that will get them there.
You really have to give OpenAI credit: its work has become excellent, and so has its community management.
Features are being brought back, rate limits are being reset, and the path forward is being communicated transparently. I use both Codex and Claude Code every day, but OpenAI has found a far better way to work with the community. Anthropic could learn from that.
@toddsaunders I don't know what's worse...the site or the fact that someone sees this and says "yeah, that's totally the service I want to use and pay for...."
Seeing people talking about token burn with GPT-5.6. Not sure what they're talking about. I've been using it like a madman today with virtually no pressure on my usage. Combine that with the resets I also have available and the hydrant is basically running wide open.
5.6 Sol is incredible by the way. Working in both Codex and Work, as well as the backend to most of my Hermes Agent fleet.....it's an absolute joy.
@OpenAI and @NousResearch are rapidly becoming significant parts of my backend infrastructure over the past couple days. I was effective before but since 5.6 dropped.....holy shit......
Already noticing a pretty significant improvement in how my main agent is communicating through Hermes Agent using Sol on high. It feels more personable and like it's actually communicating rather than...I don't know....reading from a script. Just feels more natural in the minutes/hours since the changeover.
Everybody's out here celebrating that @AnthropicAI reset Claude limits on @OpenAI day.
Of course they did.
But here's the thing: those limits are just a reset. Fable 5 is still going behind a paywall. And, GPT-5.6 will be still there.
If Anthropic is trying to be competitive today, they're going to need something better than limit resets. They need to figure out who they're trying to serve in the AI world and then create a plan around *that*. Resets on a model that will just disappear behind a paywall in a few days are not a competition structure. They're an attention grab on someone else's day.
don't really care about the gift, but the execution left a LOT to be desired. you all tried something cool here and a lot of people have had success with it....there will, of course, be those that don't go as planned.
as for the charge....looks pretty telling in my agent sessions list....
If you're a manager of a small team (retail, salon, auto, etc...) and hate making the schedule every week or two, my latest release is for you. Schedule's Up lets you use employee availability to create a schedule for the week in minutes - no app to download, no accounts to maintain. Check it out at https://t.co/RDKKM0CbPM.