Coding agents can run shell commands, install packages, call MCP tools and touch your secrets. Maybe we should watch them. 😂 Prismor is an open-source firewall for AI agents. Every tool call gets checked BEFORE it runs. Allow. Warn. Approve. Block. Codex + Hermes included.

coding agents are moving fast from prototype to production. the infrastructure question is what's left. join @parthsareen from @ollama and @zainhas Hasan from Together AI at @AIconference for a breakout on what it actually takes to build coding agents on open models, and run them at scale.

Coding agents can now create their own training environments from existing software! Xiaomi's new CodeMidas system lets agents inspect open-source projects, extract coding tasks, build execution-based tests, and discard broken or exploitable tasks. It generated 5,545 verified RL environments from 3,185 projects across 23 programming languages. According to the paper, training MiMo-V2.5 on these tasks improved issue repair by 11.7%, whole-program construction by 17%, and Terminal-Bench performance by 8.5%. Instead of humans writing every RL task by hand, CodeMidas can continuously mine existing software for new, verifiable training tasks.

Coding agents can grade their own homework. Your marketing agent can't (yet). So you're stuck babysitting it because you can't trust it. Here's how to fix this: 1. Pick one subjective deliverable. Email copy, ad creative, etc. 2. Pull 50 real examples you've shipped. 3. Have your best marketer score them. Not just "good or bad." What's right, what's wrong, and why? 4. Feed that analysis into your agent to improve to improve it. 5. Have your agent create 50 outputs based on a sample of inputs. 6. Have your marketer rate and review again. 7. Keep iterating on your agent until your expert is happy. This evals loop makes the difference between your agent giving you a sh*tty first draft that you spend more time editing than you would have spent creating it from scratch, and an AI employee who you can trust. The Verification Gap is one of 10 trends I share in my new State of AI Report. Read it here: atherial .ai/state-of-ai

Coding agents aren’t just for CODE anymore. OpenMontage turns Codex, Claude Code or Cursor into an entire video production studio. Research. Script. Images. Voice. Music. Editing. Rendering. 12 pipelines. 100+ tools. 700+ agent skills. And it’s open source. 👇🏻

Coding agents, sovereign AI, data centers, rockets, satellites. This week's biggest rounds on Funding Watch with @sophiadew: - @cognition $2B at $48B val - @MistralAI €3B at €21B val - @nscale seeking $3.5B pre-IPO - @stoke_space $1B series E - @PixxelSpace $100M series C

Coding agents can turn a vague idea into working software fast. But what happens when no one has defined what “correct” means? Markus Eisele explores how much specification agentic development actually needs:

Coding agents log what happened, but not where the time and tokens went. Here's how to get real OpenTelemetry GenAI traces out of DeepSeek Harness, viewed in Jaeger or Langfuse. 🔗 #OpenTelemetry# #Observability# #OpenSource# #AI#