Pinned Tweet
Claude Fable 5.2 has owned the timeline this week.
4,800 lines of Python → a pelican riding a bike in Blender, GUI never opened 2,000 lines of Three.js → one animation, 83 minutes of rendering A custom paint engine → 700,000 brushstrokes on a single canvas
Then the scores landed: #1 on the Artificial Analysis Intelligence Index. FrontierCode 2.0 up from 32.4% to 49.1%. Up to 30% faster output. Prices down 20%.
Code is becoming AI's universal medium. If it can be defined, executed, and checked on a computer, AI can attempt it.
Here's what the demos skip: the model writes the program. It doesn't deliver it to users. It can't open a cloud account, manage keys, set permissions, or pay for its own calls.
Code is the recipe, not the kitchen.
The bottleneck moved. It was "can we build it?" Now it's what comes after:
• which models, whose keys
• sign-up, sessions, access control
• Postgres, per-user data isolation
• file buckets and uploads
• usage, billing, a public URL
"We're just one programmer short" has become "we're just one backend short."
That's Vectoree. One project key for models, tools, auth, database, and storage.
• Model Plaza: 315 models, or just use vectoree/auto
• Tool Hub: hosted MCP tools for search and extraction
• Skill + CLI: your agent creates tables, uploads files, calls models
• OpenAI-compatible: change the baseURL, swap the key
• No monthly fee. $5 welcome credit after CLI login + link
Deploy and Payments are coming soon. We're not going to pretend they're live.
AI is learning to create worlds. We want to turn the power on.
Where do you get stuck: writing the agent, or shipping it?
OpenAI keeps a human at "review & deploy". That's not caution, it's the current boundary:
agents can investigate and propose, but they can't own the environment they change.
A project key with auth, Postgres, and storage is what moves that boundary. @OpenAIDevs
Codex already helps us build ChatGPT faster. Now it’s helping us keep ChatGPT fast.
We built the Performance Factory to coordinate agents that investigate regressions, propose fixes for engineers to review, and check if those fixes work in production.
1
7
We're one of the teams building on this. Claude writes the app — Vectoree gives the agent a backend to ship it: auth, Postgres, storage, and models on one project key.
vectoree.ai @AnthropicAI
1
17
Give your agent a backend.
Models, tools, auth, database, and storage.
One key for OpenAI, Anthropic, Google, and xAI.
Install the skill, log in, link the project.
Five dollars to start. Then pay per token.
vectoree.ai
a vibe-coded app leaking keys made the rounds again.
usually it's not a fancy exploit. it's a VITE_ prefix, or a key pasted into the frontend 'just to test'.
what's the first thing you check before you ship?
13
proactivity is only as good as the data it can reach tbh @thsottiaux
Heads up, we’re starting to test dots with a small number of ChatGPT Plus users. Looking forward to the continued feedback as we expand and I’m recently seeing a lot more positives around how good the proactivity has been after many improvements.
We’re also bringing dots to all ChatGPT Business Standard users. Rollout will start today.
10
Gpt, Terra, 6.2 @WorkBuddy_AI
1
1
9
SWE-sweep going from 39% to 12.8% just by switching harness is the real story in this table. @sundarpichai
Gemini 4 Neon: frontier-level intelligence with fast end-to-end task completion across complex software engineering, knowledge work, and science workflows.
Begins rolling out today in the Gemini API in @GoogleAIStudio, Gemini for business, for paid Google AI Pro and Ultra subscribers in Google @Antigravity, and for Google AI Pro and Ultra subscribers in the @GeminiApp.
1
2
19
so convenient @WisprFlow
Notetaker is live today on iPhone. Now, your meeting notes can go wherever work takes you, for free.
Use Notetaker on desktop for virtual meetings and on iPhone when you’re meeting in person.
Your notes stay together across devices, so you can pick up right where you left off.
8
Your agent writes code. It can't store users, or the data.
Vectoree is the backend.
Skill → login → link → gateway.
vectoree.ai
Point Codex at Vectoree in one command.
macOS / Linux:
bash <(curl -fsSL vectoree.ai/scripts/codex-ve…)
Windows PowerShell:
irm vectoree.ai/scripts/codex-ve… | iex
One project key after that.
vectoree.ai
1
13
GPT-6.2 Astra and Sol are now available. @OpenAI
Some of you have been asking very politely. Some of you have just been asking.
GPT-6.2 Astra and Sol are now available.
Super efficient and built to perform where it counts—the real world.
openai.com/index/introducing…
1
2
24
One command to connect the backend. One prompt and your agent builds the app.
Models, auth, Postgres and storage - all behind one project key. Your code stays in your repo; only the baseURL changes.
vectoree.ai
#Vectoree #AgentBackend
2
24
take this seriously is necessary @SketchesbyBoze
We have an epidemic of students turning in essays they didn’t write about books they haven’t read, and if teachers are now using AI to grade assignments en masse we may as well shut down the schools because no one here is teaching and no one is learning.
9
Most agent demos end before the interesting question: where does the agent get a fact it wasn't given?
Tool Hub puts Vectoree Search, Similarweb, Semrush and Ahrefs behind one project key, over hosted MCP.
The model writes. The tools decide what it knows.
#Vectoree #HostedMCP
1
10
The interesting thing about d1.2 isn't the index score. It's that the outputs are already shaped for an app to consume - spans, boxes, ranges, choices.
Adopting a new model shouldn't also mean adopting a new provider integration. One key, one endpoint.
#Vectoree #ModelGateway @liquidai @X
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities.
> #1 on the @ArtificialAnlys Decision Index (75)
> Read a value from the input (text, number or date)
> Find every passage of a kind in a text (spans)
> Locate things in a picture (box for one, boxes for every instance)
> Estimate a number (number with a range)
> Pick every option that applies (choice)
1/ 🧵
1
14
Anyone here need a deploy feature?
Not a checkbox - a public URL for the app your coding agent just built, so someone other than you can open it.
Deploy is in development. Reply with what you'd use it for; we're scoping it now.
#Vectoree #AgentBackend
1
11
Computer-use agents handling prior auths is the real test. The model is only half of it. The other half is auth, audit logs, and where the records live. @OpenAI
ChatGPT for Healthcare is advancing with more capable computer-use agents, stronger clinical answers, and 20+ new healthcare plugins.
@AbridgeHQ, @AmbienceAI, @Cigna, @CVSHealth, and other leading healthcare organizations are putting our models to work across prior authorizations, referrals, discharge planning, and patient outreach.
Together, these advances connect medical evidence with practical help across the tools teams already use—so they can spend less time navigating systems and more time caring for patients.
openai.com/index/advancing-c…
15















