Welcome back to Agentic Coding Weekly. Here are the updates on agentic coding tools, models, and workflows for the week of Oct 18 - 24, 2026.
1. Tool and Model Updates
MiniMax M3.1 Flash and Kimi K3.1 Pro are the new models released last week.
MiniMax M3.1 Flash is 428B total and 23B active params MoE model. Natively multimodal. Pricing is same is DeepSeek Flash models, $0.15 / $0.6 per million input / output tokens.
Kimi K3.1 Pro is number 4 on artificial analysis intelligence index, only behind opus and sonnet 5.5, and muse wonder full. Model weights expected to be released by Nov 4th.
DeepSeek launched search and fetch API. Costs $0.75 per 1000 requests. For comparison, Brave Search API costs $5 and Kagi API $12 per 1000 requests.
Here's the current state of coding benchmarks:
2. Open Source Corner
DeepInfer - DeepSeek open-sourced their inference framework
openai-cli - CLI for the OpenAI API
Corgi - multiplexer terminal agent manager inspired by herdr
Death Plan Switch - self-hosted dead man's switch for solo devs
3. Reading List
WorldTUI: I trained a 360M model to pretend to be a terminal - based on Qwen2.5-Coder-0.5B, this is very cool
Ask HN: How to deal with shipping AI-written code you don't understand? - lots of argument that this is how people at the higher level (principal engineers, architects, management) already operate, but I like this particular argument in the thread: There's a huge, insurmountable void between "I don't understand all the code" and "nobody understands all the code"...
If you found this useful, forward it to one friend/colleague who’d probably want to read it too. Got this forwarded to you? Subscribe here.
That’s it for this week. I’ll be back next Monday with the latest agentic coding updates.
— Prashant