Welcome back to Agentic Coding Weekly. This is the first issue of 2026. As expected with the holidays, it has been a quite week. This week's workflow is from Eunsu Jang on running coding agents in parallel.
Here are the updates on agentic coding tools, models, and workflows worth your attention for the week of Dec 28, 2025 - Jan 3, 2026.
Tooling and Model Updates
IQuest Labs, the AI research arm of Ubiquant (one of China's largest quant hedge funds), released a new family of coding models, IQuest-Coder-V1. The release includes 7B, 14B, and 40B sizes.
The 40B-Loop-Instruct model hits 76.2% on SWE-bench Verified after fixing some reward hacking issues in the original score calculation. Roughly in the same band as Gemini 3 Pro (76.2%) and close to Sonnet 4.5 (77.2%). Remains to be seen how it performs in practice.
Read the technical report.
Boris Cherny, creator of Claude Code, shared his setup. Highlights include keeping the setup simple, using opus 4.5 for everything, and pre-allowing common safe bash commands to avoid permission prompts instead of using YOLO mode.
Workflow of the Week
This week's workflow is from Eunsu Jang, Lead AI Application Engineer at Explaza:
Parallel Codex Agents with Git Worktrees
I use an OSS coding agent orchestrator called vibe-kanban and give it the task prompt. It creates separate git worktrees and runs independent Codex agents for each task.
If the task is genuinely hard or ambiguous, I’ll kick off multiple runs with slightly different instructions, or just switch the Codex model (for example GPT-5.2 vs GPT-5.1 Codex Max). This tends to surface different tradeoffs: one version might be safer and boring, another might be over-engineered, another might match the existing architecture better. Since each implementation is isolated in worktrees, it’s easy to compare approaches without branches getting tangled.
I noticed that these models often over-engineer the implementation. So I add following instructions in the project AGENTS.md file:
## GENERAL RULES:
- Avoid over-engineering
- Follow DRY/YAGNI
- Respect the project architecture
- Fail fast. No invisible fallbacks.Once the agents finish, I deploy each worktree and test in the browser. Anything that fails basic self-QA gets dropped early. For the ones that look promising, I open a separate Codex session and have it score and compare them. The score is subjective, but it’s useful for forcing a clear decision.
Here’s the exact comparison prompt I use:
Worktree A and Worktree B are results of the task below. Please evaluate each on a 10-point scale based on:
1. No over-engineering
2. Proper use of DRY/YAGNI
3. Strict respect for the project architecture
4. Fail fast, no invisible fallbacks
Choose the better worktree and explain why.
Original task instruction:
...Finally, I select the winning worktree, issue small follow-up instructions for final polish, and create the PR.
— Eunsu
I write this weekly on Mondays. If this was useful, subscribe below:
Community Picks
2025: The year in LLMs
Year-end review from Simon Willison. Good overview of how the landscape shifted over the past year. The sections "The year of coding agents and Claude Code", "The year of vibe coding", and "The year of programming on my phone" are relevant for agentic coding. Read more.
AI Is Forcing Us To Write Good Code
100% test coverage, clear docs, small modules, static typing, fast dev environments. The things we used to cut under time pressure are exactly what makes coding agents effective in a codebase. Read more.
Vibe coding a bookshelf with Claude Code
Author used Claude Code to build a digital catalog of his 470 physical books.
Workflow: photograph books, extract metadata via OpenAI Vision API, manually validate JSON, fetch covers via Open Library and SerpAPI, then use Claude Code to build the UI.
It's a great look at how coding agents help us build personal tools and projects that we'd normally skip because they're too much hassle to set up.
The Future of Software Development is Software Developers
Jason Gorman argues AI isn’t the end of programmers. Current hype surrounding LLMs replacing developers is merely the latest iteration of recurring technological cycles (e.g., 4GLs, Visual Basic) that failed to eliminate the need for programmers.
Key quote: "Edgar Dijkstra called it nearly 50 years ago: we will never be programming in English, or French, or Spanish. Natural languages have not evolved to be precise enough and unambiguous enough. Semantic ambiguity and language entropy will always defeat this ambition."
Ask HN: Any example of successful vibe-coded product?
HN thread where people share products they've built primarily using agentic coding. Common pattern: building tools for themselves or people they know instead of paying for SaaS. Read HN comments.
That’s it for this week. I’ll be back next Monday with the latest agentic coding updates.
