Intelligent routing across every single AI model (GPT, Claude, Gemini, Deepseek, Hugging Face, etc). Deterministic, secure, enterprise-ready.
- Tweets 687
- Following 95
- Followers 71
- Likes 107
Treat streaming as a sequence of events, not a successful request followed by some text.
Track completion, read the stop reason, and handle errors inside the stream. A response that hits its token limit needs different handling from a finished answer.
1
3
Test the interruption deliberately.
Cut the connection halfway through a response. Check that partial output is marked incomplete and downstream actions don't run on it.
Before retrying, check whether any earlier tool actions already happened.
1
3
Test this workflow on Gatewayz.ai today!
2
Your AI app starts writing an answer. Halfway through, the connection drops.
The user sees text. Your backend may already have logged a successful HTTP response.
Does your app know the answer never finished?
1
6
OpenAI's dots launch puts ongoing work at the center of the agent conversation.
A useful test: give an agent a responsibility for a week.
Can you see what it did, correct it halfway through, and trust the next run to use that correction?
Introducing dots, powered by GPT-6 Astra.
Remarkably capable, always-on agents built to handle everything.
28
One detail in the Sonnet 5.5 launch deserves attention: Max effort scored below Xhigh on FrontierCode.
Anthropic says some runs timed out or made changes outside the task's scope.
Finishing the requested job matters more than doing extra work.
Introducing Claude Sonnet 5.5, the second model in the Claude 5.5 family.
It’s a clear upgrade over Sonnet 5, runs more than 30% faster, and costs up to 30% less for most work.
3
64
Weekend build idea: a dependency-update brief.
Give an agent the libraries your app uses. Have it read new release notes and flag changes that might affect your code.
Start with a draft for a developer to review.
1
1
37
Each finding should include the installed version, the proposed version, the relevant release-note passage and the files worth checking.
If the agent can't establish whether a change applies, mark it unknown.
1
9
"Find five good examples" and "find every qualifying company" are very different jobs.
Exa's Agent Ultra is aimed at the second.
For builders, the interesting question is how you measure what's missing from a research agent's answer.
Introducing Agent Ultra - a step change in deep research
Agent Ultra orchestrates swarms of agents to perform exhaustive research, build comprehensive lists, and answer questions requiring thousands of sources.
In both evals and vibes, it's state of the art
1
27
Give each agent its own API key.
Gatewayz lets you set a request cap per key, so a research experiment doesn't need the same allowance as a production workflow.
When a cap is reached, handle the error and pause the job.
1
26
Set up a key for your next agent experiment:
Gatewayz.ai
9
Build idea: a meeting assistant that keeps track of who promised what.
Start with speaker diarization alongside transcription, so each commitment can be traced to a speaker turn and timestamp.
1
22
Add a review step before creating tasks or sending follow-ups.
Speaker labels aren't verified names. Let people confirm who's who and what they agreed to.
The useful output is a checked action list, not just a shorter transcript.
1
6
Try building something similar today at Gatewayz.ai
3
A typo in a model name should be a fixable error.
Gatewayz returns 400 model_not_found for an unknown model ID, giving your agent code a clear reason to stop and fix the request.
Retry logic should distinguish a bad request from a temporary outage.
2
44
Python Workers going GA gives Python builders another route from prototype to deployed app.
A useful first project: a document assistant that extracts fields, flags missing information and prepares a draft for review.
Start with one workflow people already need.
We're excited to announce that today Python Workers are generally available (GA).
It means Python is now a first-class, fully supported language on the Cloudflare Developer Platform. Read everything about it: cfl.re/4he07Gh
27
Gatewayz AI retweeted
I gave Jev the keys to Arthenix city.
Now it runs 124 lives.
Each has their own personality, job, habits and friends, and Jev decides what they do: when they work, when they take a break, who they hang out with, when they head home.
A truly living world. 🧵
5
4
43
24,740
