Don't wait until production goes down in the middle of the night. Shift left to catch the problems as they're created. Sam Anthony shows you how to use AI agents to bring testing, security, compatibility, and deployment validation earlier into CI/CD. ⤵️
IBM Developer
IT Services and IT Consulting
New York, NY 59,611 followers
Welcome to a global community of builders, thinkers, and problem-solvers.
About us
Join a global community of builders and problem-solvers. IBM Developer is your space to explore the latest in AI and DevOps through tutorials and expert insights
- Website
-
https://ibm.biz/BdepWm
External link for IBM Developer
- Industry
- IT Services and IT Consulting
- Company size
- 10,001+ employees
- Headquarters
- New York, NY
- Founded
- 1911
- Specialties
- developers, cloud, artificial intelligence, blockchain, nodejs, Swift, Data science, AI, and serverless
Updates
-
How much does the agent harness matter when the model stays the same? We ran CUGA, Hermes, and Deep Agents on the same 90 AppWorld tasks with GLM-5.3-flash. Depending on the setup, the agents completed anywhere from 30% to 81% of the tasks. See how we configured each harness, then explore what a harness can control at runtime. In CUGA’s Policy Playground, give two agents the same request, “Pay everything I owe,” and see how a policy changes what gets paid. Read the breakdown and try the demo → https://ibm.co/6041EnH7t
-
-
Want to use IBM Bob in an editor you already work in? With ACP (Agent Client Protocol), you can work with Bob in compatible clients like IntelliJ and Zed. The protocol defines how prompts, session updates, tool-call activity and permission requests pass between Bob and the client. Bob reports what it’s doing, while the client handles how it’s displayed, giving you more choice in where you work. Explore the protocol → https://ibm.co/6042EV7cO Get started with Bob → https://ibm.co/6043EV7cP
-
IBM Developer reposted this
Which model should I use for this task? Bob takes that decision off your plate. It works with multiple models, each with different strengths, speeds and costs. The model router classifies your first request and selects a model for the task, balancing quality and cost across Fast, Premium and Ultra tiers. Here's how a model earns its place in our routing table: • Open benchmarks and BobBench, built around real enterprise use cases • Prompts and an agent harness tuned for each model • Randomized A/B testing with internal IBM users • A rollout based on the results With 170,000 IBMers using Bob, we learn from a substantial base of real enterprise work. Our new blog post walks through the whole process, including why Bob keeps a task on the same model once it's routed. https://lnkd.in/ezX_Q98V #IBMBob #AI
-
IBM Developer reposted this
-
The unofficial #IBMTechXchange numbers are in. Which stat belongs in your end-of-week report?
-
IBM Developer reposted this
Have an idea that could make a difference? Bring it to life with the AI Builders Challenge North America! Build a solution through AI for Good, tackling real-world needs like education, public health, and community resources. Or explore AI for Business and the Wildcard Challenge. College and university students can compete for $15,000 in prizes with no coding experience required! 🗓️ November 2–December 10 | Register now: https://bit.ly/4rRaEeb #aibuilderschallenge #ibmskillsbuild #studenthackathon #ibm
-
-
IBM Developer reposted this
AI in Action | 2026 Edition 75: IBM TechXchange 2026 Want to know the major talking points when thousands of devs, engineers, architects and researchers gather to build with the latest tech? Read now and subscribe ⤵