Here’s a great example of AI agents allowing you to do new work that wouldn’t have happened before. Today I was in a meeting with a customer where we demoed a product experience using a full branded experience on their end. The app was built primarily using AI, and allowed us to show the full value proposition much more realistically. Before AI this would have been done with slides or a far more generic demo. Which obviously means the customer does far more work to imagine the value proposition, putting the onus more on them. But with AI, it became cost effective (time wise) to actually just build the customer a working prototype. This will likely change our complex sales processes going forward. When we talk about AI agents allowing you to do more things than you did before -not just replace what we already do- this is a perfect example.
How People Are Using AI For Coding
In 2025, professional developers pair tightly with AI to ship production work: multiplayer backends arrive via Cursor and Grok in hours; Claude Code can dump ~20k LOC in a 2.5-hour run that still needs review; teams route output through guardrails like Codacy’s MCP server and terminal agents such as Gemini CLI (1M-token context, rate-limited), with editors moving to AI modes (VS Code). The working rhythm tilts to “tight-leash” AI development—load context, ask for options, draft, review docs, test, commit—with model swaps for stability (e.g., Sonnet → Gemini 2.5), error audits (Chip Huyen: “content not found” 20–30%), and repo restructuring to cut agent steps (8→7). Organizations report scale effects (claims of >⅓ code generated at Google), while leaders emphasize speed and cross-language execution (Andrew Ng). Agents also act beyond code (sending emails or automating flows), pushing teams to enforce explicit permissions, CI checks, and human review before merge.
🤖 ai summary based on 25 tweets
Stories from AI users. WhatsNextForAI curates public sources. No product affiliations. Opinions are the authors.
Can’t load? Read tweets in preview...
Ok finally fixed it, so ChatGPT helped me write code that 1) in the first few minutes of loading, checks the DOSBox <canvas> for a lot of blue pixels, then if there are: 2) loads Tesseract.JS, a JavaScript OCR library, and reads the text on the screen 3) if it says "Welcome to DOSBox" it knows it's F'd so it reloads the page 4) in most cases reloading the page fixes the bug and Windows 3.11 loads properly!
I asked Claude Code "can you write code to make you email as me" and it did in less than a minute (in Python) And then it sent me an email introducing itself! Now I can let it send annoying emails to the local municipality about holes in the road like I sent today myself, just by asking it! That \! is a bug but it'll fix it, and inb4 it only has sending (SMTP) access of course 😊👌
An attempt at single shot having Claude Code build a Sentry clone. It failed but it wrote 20118 lines of code without interruption for 2.5 hours. At one point it got low on context. Was using a ton of sub agents. It's not there yet, but man it's getting close. https://t.co/IEFCC1EXHh
I stopped using sonnet. It felt like coding with a drunk friend—fun until he installs new packages or edits other components “for fun.” My prompt used to be 50% instructions and 50% dont-dos. Switched to gemini-2.5. Finally, some peace. https://t.co/9aPyBOxq8C
IT WORKS!!!!! A FULL multiplayer with Python websockets server that receives and broadcasts all player positions every 100ms (10 times per second) All code written almost 100% by AI with Cursor and Grok 3 wrote the server code Now you can fly around with everyone else :D It'll probably crash after this tweet but let's try! https://t.co/UVJczDtWZ5 The reason before it didn't move was it was stale players that already quit and it kept their positions Every red plane is someone real flying now too! 😊