Based on what? Each model so far has been noticeably better than the last, so I don’t see why the next wouldn’t be too?

"Can’t you just draw an exponential line on the curve?" - Dario Amodei, February 13th 2026.

But this time draw it for spending expectations.

Are you predicting that OpenAI is out of business in 6-12 months?

I'm not an AI booster, and I don't see "sustainable" in the current markets, but I'd take the other side of that bet!

Even if that were true, it wouldn’t make the growth exponential. There are many other trajectories out there.

Even if it’s not exponential, I still can’t see logic in it going to zero.

Still a fraction of the cost of most other providers, and wouldn't shock me if we see the others all doing something similar.

Exactly the same as the OpenCode drama, at this point OpenAI are just getting free customers from common sense.

Anthropic seem oddly focused on their moat being the app, rather than the underlying intelligence and model. Which is odd when Claude Claude is no better than all the other harnesses.

I have a suspicion that the reason they’re using react, and the reason they’re trying to keep people using Claude code instead of other tools is because it gives them additional data to train on.

It's even simpler than that. They keep trying to get people using Claude code because that's their vendor lock in. That's their business model. They don't want to become a commodity, they want you to use Claude Code. I can't blame them either.

1-2 engineers at Anthropic can implement their own OpenClaw in a matter of days (and give it a less stupid name while they're at it).

This is just a marketing move by OpenAI.

Even the thousands of integrations?

The strenght of openclaw is the massive community adherence.

It can connect to a ton of things.

It's like the https://zapier.com meets LLMs.

I'm sure a lot of API plumbing can be copied/adapted wholesale from the (open-source) openclaw repo. LLMs are surprisingly good at this kind of stuff. And yeah it would require some testing, but I doubt what openclaw has now is itself in a very stable state (from my very limited testing)

You think he implemented those thousands of integrations himself? Or maybe some particular tool was used that can be used again for implementing such things? Particular tool that so many of us use as well?

It was obviously largely by the community. Just take a look at PRs.

Question is, will the community continue with this level of engagement on the project now that it has OpenAI stamp on it?

I highly doubt. A fork? Maybe.

Maybe they don't want the idiots' money.

Honestly the trademark thing was fine on its own — you protect your brand. But Anthropic was playing a legal game while OpenAI was playing a relationship game. One got a name change, the other got the most visible open-source agent project. If you're going to push someone away, maybe think about where they land.

Miss the boat on what? Idiotic agents with no sandboxing, with the door fully open for prompt injection purgatory? There's nothing interesting about this technology other than how dumb it is and how badly it will spam the internet (see https://theshamblog.com/an-ai-agent-published-a-hit-piece-on... for how bad agent slop has gotten).

Interesting that it’s seasonal, I guess tax season?

Yes. Tax season for small businesses and New Year's resolutions for personal finances are the times I get the largest influx of new signups and subscriptions.

Not sure I’d be worried for my job, but it’s legitimately a significant jump in capabilities, even if other models attempt to fudge higher bench results

Also wondering, why not just use DPDK?

First, "just use" is doing a lot of work in that sentence, because DPDK is much harder to use than XDP. The authors of this blog were surprised they had to do their own checksumming, for instance.

Maybe more importantly: they're not building a middlebox. DPDK ultra-high performance comes in part from polling. It's always running. XDP is just an extension to the existing network driver.

I think they are getting a lot of value from the rest of the Kernel's networking (VETH/namespaces etc talking to containers).

XDP is built into the kernel. DPDK is a huge framework that invasively bypasses the kernel and has to remain compatible as an external project.

They're almost never straight upgrades for the exact same prompts across the board at the same latency and price. The last time that happened was already a year ago, with 3.5 Sonnet.

Wow, they are sunsetting all models after the launch of GPT-5 - Bold statement.

Same background as me, however I moved into cellular/mobile - not really any official routes I know of in the industry, certainly not like a Cisco track.

Typical route is work at a Telco or IoT company as Network Eng or Developer and naturally pivot into telco learning on the job.

Vendors will run training courses when you buy their kit which helps a little, but it’s mostly self learning or on the job.