Phi-4 (14B) beats GPT-4o on MATH. 80.4% vs 74.6%. Thousands of times cheaper.
The question isn't which frontier model to pick. It's what domain fine-tune and orchestration layer you can build that a generic API can't touch.
Best data + right tooling + small model = moat.
OpenAI's Sol/Terra/Luna tiers sound like cost savings. They're not. Optimizing routing across three price surfaces locks you into one vendor's architecture. When they change prices (114 models changed in March), you're rebuilding your cost layer. Route across providers, not tiers.
Codex is dead. Atlas browser is shutting down. Both absorbed by the platform.
The pattern: if a platform can ship your product in 6 months, you're a feature.
Cursor survived. Codex didn't. The difference is where the data accumulates.
3 companies took 67% of AI VC in Q1 2026. GPT-5.6 needed a government preview. Fable 5 got pulled globally for 19 days. Open weights can't be revoked. That's the only fair fight left.
The new ChatGPT app merged Codex in and made it the primary interface. Chat is now secondary. I hate it. The website is literally better at this point. Sometimes I want to chat, sometimes I want to code. Claude keeps these separate and it works. Am I the only one?
The base model layer is commoditized. LongCat-2.0 beat GPT-5.5 on SWE-bench Pro for 2 months under a fake name. Nobody noticed. The moat is no longer the model. It's the workflow, the data, and who can't afford to switch.
7,000 Meta engineers got reassigned to AI teams they didn't choose. Zuckerberg just admitted the bet 'hasn't progressed as quickly as expected.' The hidden risk of a big tech job right now is real. At a startup, at least the failure mode is yours.
114 of 483 AI API models changed prices in March alone. OpenAI called their own pricing "accidental." Claude Sonnet 5 jumps 50% after August. If your margin depends on token costs staying fixed, open-source isn't ideology. It's unit economics.
SaaS killed enterprise software licenses in the 2000s. AI agents are doing the same to SaaS now. $285B wiped from market caps. Seat pricing fell from 21% to 15% in 12 months. Price for outcomes, not access. The playbook just got rewritten.
Gartner found ~130 vendors with real agentic AI. Thousands are selling it. The rest is RPA with a rebrand. Real test: give it something it wasn't trained for. Real agents adapt. Washed agents break.
Tokenmaxxing was always a vanity metric. Uber burned its entire 2026 token budget in 4 months. Meta pulled its leaderboard. Microsoft cancelled subscriptions. Companies are finally asking what AI is actually producing, not just consuming.
OpenAI's Daybreak launch is a signal. Enterprise is 40%+ of revenue, heading to parity with consumer by year end. ChatGPT has 900M weekly users and is still losing share. Consumer ceiling is real. Enterprise vertical moats are where the durable value goes.
Two frontier model launches. Two weeks. Both restricted or killed by government order. OpenAI caved June 26. Anthropic's died June 13. Closed APIs now have a control layer above the vendor. Open weights don't. Not politics. Architecture.
70-95% of AI agents fail in production. 88% never reach it. 'Agentic' and 'autonomous' are not the same thing. Most vendors conflate them on purpose. Human-in-the-loop is not a bug. It's the architecture.
Linux beat Windows Server. Apache beat IIS. Android beat iOS globally. Open-source AI will do to Anthropic and OpenAI what the open web did to AOL. Qwen just hit 1B HuggingFace downloads. The infrastructure layer always goes open.
Anthropic killed Fable 5 and Mythos 5 for every enterprise customer 3 days after launch. Stripe, Mozilla, Amazon Bedrock. All down. Overnight. If your product runs on one model, you don't have a product. You have a dependency.
I went to In-N-Out and ordered a cheeseburger. The cashier, a calm young woman named Destiny, asked me a question I did not expect.
"You want that Animal Style?"
I paused.
I did not know what this meant. But a samurai does not admit he does not know. So I answered with weight.
"...Animal Style."
"Cool. So that's mustard-grilled, extra spread, grilled onions, pickles. Yeah?"
I understood now. This was a sacred permission. For one meal, I was being told to put down my manners at the door. To eat the way a beast eats, without shame. I had waited my whole life for someone to give me this order.
"Yes," I said. "I will become the animal."
Destiny did not blink. "...Okay. You want your fries Animal Style too?"
I stopped. Even the potatoes?
"The potatoes also become animals?"
"I mean, they get cheese and sauce and grilled onions, so..."
"Then yes. Let the potatoes abandon their restraint as well."
"...Got it." She was the calmest woman I have ever met. "3x3, 4x4, or just the one?"
I did not know these numbers, but I knew a challenge when I heard one. "How many must I face?"
"It's, like, how many patties you want."
"How many is the most honorable?"
"...Four is a lot."
"Then four. A warrior does not ask for fewer."
She wrote it down without argument. A 4x4, Animal Style, with animal fries. She warned me once, kindly. "That's gonna be huge." I told her I was counting on it.
It arrived. It was a tower. Cheese and sauce ran down my hands the moment I lifted it. There was no clean way to eat it. There was no dignified way. That was the entire point.
I ate it like a beast. Both hands, no honor, grilled onion on my chin, and I have to be honest with you, it was the best thing I have ever put in my mouth.
For thirty years I have kept my manners at every table in the world.
They handed me a burger and told me to be an animal, and I have never felt so free.
So tell me, America.
The whole country knows the secret menu. What else are you hiding in plain sight?
And "Animal Style." Was I eating the animal, or finally becoming one?
At WWDC 2003, someone shouted, 'Take out debugging!' at Steve Jobs mid-demo. Everyone laughed. It was absurd. In 2026, AI agents read your errors and rewrite the code. The heckler wasn't wrong.
https://t.co/HCNtcUnCZA
Founders budget AI like SaaS. Agents bill like AWS. Anthropic pulled a billing overhaul on the day it launched. Also being sued for delivering 35% of advertised usage. Mid-tier agentic costs: $3-15k/month. Open source sidesteps all of it.