Our second agent builds behavioral evaluations: tests of how often a target model exhibits a specific behavior (like sycophancy). Our agent designs, codes, runs, and analyzes evals. They consistently work: 88% of our agent’s evals measure what they’re supposed to.
AUTOMATION
-
Boston to NYC: Ultra-productive commuting with mobile work
By
–
Breakfast in Boston, working Lunch in NYC, back home by dinner, while working on the train in both directions. Some people would even commute that way.
-
The shift from manual development to AI-driven app creation
By
–
Before: Hire developers ($10K+), wait weeks, hope it works, pay separately for hosting, database, auth, etc. Now: Describe your idea, watch it build, deploy instantly, everything included – a ready-to-go app in less than 10 minutes
-
Using AI agents to build custom business management applications
By
–
Someone told Mocha, "I'm a restaurant owner and I want to build an app that tracks all my expenses, vendors, daily profit and loss, and growth projections."
— God of Prompt (@godofprompt) 24 juillet 2025
I simply entered this prompt and let Mocha work its magic: pic.twitter.com/AozGIYpWHpSomeone told Mocha, "I'm a restaurant owner and I want to build an app that tracks all my expenses, vendors, daily profit and loss, and growth projections." I simply entered this prompt and let Mocha work its magic:
-
Mocha AI tool automates end-to-end software development
By
–
I'm scared.
— God of Prompt (@godofprompt) 24 juillet 2025
This AI just made devs optional.
Mocha can:
• Build an app
• Add Gmail login
• Hook up a DB
• Deploy the backend
• Go live on a custom domain
All from a single text prompt.
Here's how it works 👇 pic.twitter.com/iITF1Ni9XuI'm scared. This AI just made devs optional. Mocha can: • Build an app
• Add Gmail login
• Hook up a DB
• Deploy the backend
• Go live on a custom domain All from a single text prompt. Here's how it works -
Piloting Generative AI in Enterprise: Governance and Implementation
By
–
𝐋𝐞ç𝐨𝐧 𝟑/𝐄𝐩𝐢𝐬𝐨𝐝𝐞 𝟐 -𝐘 𝐚-𝐭-𝐢𝐥 𝐮𝐧 𝐩𝐢𝐥𝐨𝐭𝐞 𝐝𝐚𝐧𝐬 𝐯𝐨𝐭𝐫𝐞 𝐈𝐀 ?
Mardi, on parlait des IA qui vrillent.
Aujourd’hui, on passe au test terrain : comment encadrer une IA générative en entreprise ? Et on prend une bouffée d’inspiration avec un cas -
AI Summarizes Meetings with Speaker and Slide Details
By
–
Memory AI can process your recorded meetings and recognise who said what and summarise the context from your recordings, including visual details like presented slides, for example. This is a huge enabler for those who spend their days in meetings at work!
-
Industrial Robotics Builds Manufacturing Resilience and Scale
By
–
See how industrial robotics is helping manufacturers build resilience and scale: https://
buff.ly/UCcyuEE #Roots_iiot #DigitalTransformation #IIoT #ManufacturingInnovation #IndustrialAutomation @IIoT_World via @fogoros -
AI-Generated Influencers: Ethics Behind Engineered Personalities
By
–
We’ve had AI-generated influencers for years. I’ve always hated them. Not because they’re AI, but because I always think of the human(s) behind the screen and get creeped out that they took the time to engineer a personality that isn’t theirs to better “manipulate” (even in a
-
Tool Use in Agents: Managing Complexity and Forcing Tool Selection
By
–
2. Tool use Given how easy it is to add new tools (e.g., with MCP servers), the number of tools a user adds to an agent can explode. Too many tools make it easier for the agent to choose the wrong action, making them dumber. They caution against removing tools mid-iteration. Instead, you can force an agent to choose certain tools with response prefilling. Ex: starting your response with <|im_start|>assistant<tool_call>{"name": “browser_ forces the agent to choose a browser. Name your tools so that related tools have the same prefix. Eg: browser tools should start with `browser_`, and command line tools should start with `shell_`