Agentic Coding Weekly’s Post

This already looks very slick for multi-session multi-agentic development. I'm thinking more slickness and performance improvements are in the way, plus the ability to re-arrange or split and redock the windows to one's preferences will be made available.

DEAD LINE on ScriptLix gets one crucial thing right about real time thrillers. The clock only matters if the screenplay keeps translating time into shrinking options. The setup is clean. A systems administrator in Moscow copies proof of an election interference operation, runs into the Metro, and tries to reach the American Embassy before the state closes every exit around him. That premise would be easy to oversell. The script avoids that trap by staying concrete. Battery percentage, station spacing, tunnel routes, winter exposure, and surveillance coverage do the heavy lifting. The pressure never feels generic because the screenplay keeps converting minutes into tactical losses. That discipline is what makes the concept produce pages rather than just a pitch. The audience can feel the field of possibility narrowing. A phone is not just a prop. It is evidence storage, navigation tool, liability, and countdown clock at the same time. The best real time scripts understand that compression works only when every object carries multiple jobs. From a development perspective, that is the encouraging part. Constraint becomes a value multiplier when the writer knows how to make it structural rather than decorative. Which real time thriller do you think uses one ordinary object most effectively as both practical tool and dramatic clock?

Warp is an agentic development environment, born out of the terminal. Use Warp's built-in coding agent, or bring your own CLI agent (Claude Code, Codex, Gemini CLI, and others). https://lnkd.in/g3Xu7twm

https://lnkd.in/gZSiinmh This is a great example of why documentation matters before things get too big to explain. When the people who built the system leave, the knowledge leaves with them unless someone captured it. This is the problem technical writers solve before it gets expensive.

Wouldn't it be useful if you can apply edits in your code to multiple lines, columns or even specific parts of your code at once? In this video Jesper shows various ways to block edit your code in CLion: https://lnkd.in/eu3hDcHG #CLion #IDE #QtDev #Cpp #CPlusPlus

There's a guy on X claiming he runs 100 parallel coding agents at once. Is that real? Is anyone actually shipping code that way? We put the question to 7 of our senior engineers, the ones who actually build coding agents for a living. The answer is more useful than the hot takes. Most of them run 2 to 4 foreground agents at a time. Mark runs 20+, but most of those are background jobs that ship a PR and wait for him to merge later. Igor caps at three because output quality drops fast on complex work. Florian and Imanol both plan with one model and execute with another. The bottleneck has shifted from writing code to verifying it. Everything else is theater. Featuring Mark IJbema, Evgeny Shurakov, John Fawcett, Igor Šćekić, Florian Hines, Kirill Kalishev, and Imanol Maiztegui. Full post: https://kilo.codes/S5ZxoAs

Developers usually ask one thing after every bug screenshot 👀 Can you share the complete context? And honestly, that’s where QA reporting starts getting messy. A single screenshot rarely explains the issue properly. Too many screenshots make things harder to follow. And long screen recordings? Most people don’t even watch till the end. This becomes even more painful during: → regression testing → UI comparison → production bug reporting → checking behavior across multiple windows I recently tried “Screenshot with URL” by SelectorHub and found a few features that actually feel useful in day-to-day QA work. What I liked: ✔ Full-page auto scroll capture ✔ Dual Select to compare two sections together ✔ Split View for handling multiple browser windows ✔ Selected area capture ✔ Screen recording ✔ Built-in editor ✔ Network & Console logs The Dual Select and Split View features genuinely stood out for me because they solve a very common reporting issue testers deal with regularly. Little workflow improvements like these save a lot of back-and-forth between testers and developers during debugging. Good to see tools being built with practical QA problems in mind 👏 Thanks Sanjay Kumar and the SelectorHub team for building tools around actual QA pain points It’s Free to explore and use : https://lnkd.in/gKeYv99n

Recently, I noticed something had changed in Anthropic/Claude. In the past week, my usage was being used up very quickly. On their pricing page, they note that Opus 4.7 uses ~35% more tokens. But it was more than that, and I can't really explain why - and all else equal I'd prefer transparency than not.... however I get this is a fast moving market! I am looking at a number of things, including coding locally or with other models. But the first thing spend some time tonight changing commands, because now that tokens are running out quickly, I need to think about efficiency. So a number of changes have been made including planning with Opus and coding with Sonnet. Who knows how this all goes! Has anyone else noticed anything different in the past week?

Gemini 3.5 Flash is now available in Opper 🔥 Agents need models that can keep up: plan the next step, call the right tool, coordinate with other agents, and keep going until the job is done. That’s why we’re excited to add support for Gemini 3.5 Flash in Opper. It works across the Playground, API, Agent SDK, and your coding agents. One line of code. Gemini 3.5 Flash everywhere you build with Opper.

3 projects open. Claude Code and Codex running on each. Less than 500MB of RAM. That's Zed in my current setup. For context: I canceled Cursor Pro a while back and made Antigravity my main IDE. I'm not writing much code these days, the agents do most of it, but having an IDE to participate in and audit the work matters. The problem is my machine predates the era where RAM became a serious commodity. And modern agentic IDEs are heavy. Antigravity with multiple projects open, each running its own coding agent, easily crosses 10GB. So I dusted off Zed. Reproduced the same workflow. The number above is what I got. The shift in our role as engineers is real: we're auditing more than typing. But the tools we use to audit shouldn't cost more resources than the work itself. If you're on constrained hardware and running multiple agents in parallel, Zed is worth a serious look. #Zed #AgenticCoding #DeveloperTools #ClaudeCode https://zed.dev/