DeepSeek Harness using DeepSeek Flash API

Nothing else, DS is smart enough to do everything I ask for and more, never hallucinates and never disappoints. Apps in Rust, Swift, Node, we even designed a new language better suited for LLMs, a compiler, and it runs beautifully. Haven't coded a single line since May 2026 and I will remember that date as the day AI took over my mundane job as programmer and promoted me to architect or director of my own apps

A custom plugin to ring a bell whenever a task is finished and a gong when a question needs an answer, that way I can play backgammon or mahjong in the meantime

I can't be happier

1) none

2) Local LLMs on an M1 Max (Unsloth, LM Studio, occasionally direct llama.cpp). Muse Glimmer (because it is good at what it is optimised for: interactive coding sessions). Qwen 3.8 27B occasionally and a couple of other things for fun (an LFM MoE, experimenting with Nemotron Bolt 74B-9B, enjoying the really super Gemma 4 12B QAT when working on creative photography research or non-code planning).

No openrouter (yet) but that will definitely happen.

3) Currently focusing my attentions on Pi via Paseo (and trying Pendant). Pi is just much better in smaller context sizes, and it is the right fit if you are keen to learn from first principles.

I tried OpenCode first. I couldn’t quite get on with it and what ultimately made me switch was trying to understand its terribly messy source code.

4, 5) I am mostly using AI as a coding sidekick so I don’t really have much need for orchestration yet. I did some experiments with Osaurus (really quite impressive) and I actually think that is almost enough by itself, except that I want a little more control over the sandbox container.

I feel that my usage pattern is still settling, because I am not keen to surrender all of my coding to the machine. But I am trying more purely “directed” coding to learn the process, and finally beginning to get a feel for how I will work with that.

- Antigravity (AGY - Google Ultra $100/month), mostly gemeni flash but I benchmark Gemini Astra sometimes. My main driver because it’s pretty good at search compared to setting something up with DSH.

- DeepSeekHarness (DSH) Desktop using:

—- with pay as you go DeepSeek tokens, more expensive to use during my night/Chinese daytime, otherwise very cheap

—- oMLX on my MacBook driving mostly a Qwen 3.6 35b A3b MOE (mostly used for benchmarking a less capable model)

I mostly do custom subagent programming and can trigger subagents in AGY and DSH easily enough. DSH supports a Python coordinator orchestrating DSH subagents as conversations via RPC. AGY only supports a sub-agent orchestrating other sub-agents but it works well enough at this point. I plan on releasing my subagent framework soon (called autofab in pypi), which is basically a specification driven code development agent system (using existing agent harnesses rather than implementing my own).

I tried goose, OpenClaw, Claude, codex, etc…and they didn’t really work for what I needed. I might get a $100/month ChatGPT subscription to go with my Google one, primarily to get a different perspective and benchmark openAI models on codex (Claude doesn’t really support what I need it to, otherwise would go for that instead).

I use the Local harness in vs code with a GitHub copilot max plan. Gives access to the best models from OpenAI and Anthropic (I use 6.1 sol, Astra, and opus 5.5 for most things). The Local harness has granular keep/undo and nice interface in general for reviewing changes, which is important for mission critical code. Unfortunately the copilot team is trying to force the less capable, more YOLO-mode "copilot" harness on people going forward. The community is somewhat getting them to walk that back but they seem to be missing the mark on what differentiates their product until people complain.

I also use chatgpt desktop for some tasks through a chatgpt teams/business plan. I briefly tried to set up codex cloud and it was a mess, not following startup instructions so I gave up.

1. Claude $200 + Codex $100. I never hit limits since realizing that trying to have too many agents running at once just wasn't effective. I'm probably overpaying.

2. No

3. Lynx (lynx.railcode.dev) & CLIs (less and less since Lynx)

4. No

5. Tried a few but never stuck around

I use pi as a harness, and Paseo as a Mac UI (because I need a Ui).

I tried a bunch of models, but so far deepseek flash from October is just dirt cheap; and roughly on par with the frontier open weight models.

It’s still somewhat behind closed weight models (I only know Gemini Argon) in that it produces lots of text that is approximately correct —- frontier models seem to be very good at producing concise text that’s very correct.

I pay per token, using an EU hosting provider (melious.ai).

But I can fill in the gaps myself, so for me developing for fun, this is enough.

I use AI primarily through work, I don't pay for a personal AI sub.

I do have some OpenRouter credits but I mostly use them sparingly to try open weight models out for small personal projects.

I currently interact primarily with Claude Code, but most of my async workers are on Codex.

The orchestration part is custom. I talk to a "chief of staff" Opus instance to ingest various workstreams (Jira tickets, alerts, personal ideas, etc) and turn them into beads. It launches workers to tackle individual beads using Remuda, an agent launcher I wrote at work in 2025 and open sourced earlier this year: https://github.com/yendo-eng/remuda. I configure profiles in Remuda that launch containerized agents using my preferred harness and model. I update these profiles as the pareto frontier shifts. Currently it's GPT 6 Luna max for most tasks.

It monitors the workers and has them go through at least one round of agentic code review. It surfaces PRs for me to read once they've passed agent review. It then either ships them or hands them off to my coworkers for codebases where I'm not allowed to just merge PRs.

Around this workflow I had it write a small collection of skills and scripts to grease the wheels and make repetitive actions more reliable.

Since Remuda supports several terminal multiplexers, I've swapped out the backend 3 times: tmux -> herdr-tui -> herdr-core. There's experimental support for Rex but I'm waiting for it to mature a bit before I try and daily drive it.

It's not nearly as polished as some of the software factory setups and fancy GUI orchestrators I'm seeing, but it's light, hackable, and so far the bitter lesson has kept this working really well for me so I haven't felt much of a need to try anything else.

I have been trying out Muse for a while but it hasn't found a solid place in my life yet. Still too unreliable and too many privacy concerns to justify giving it a lot of access.

I use the Zed IDE with Opus 5.5 pro subscription. I use to have access to Fable with the max plan and I miss it, despite what everyone is saying I still find that Opus does weird things. Just yesterday I pointed it to a queue suddenly being twice slower to be processed in prod. A select SQL query called thousands of times per minute became slower because of a data increased. Opus proposed several non trivial ways to optimize the query by forcing some execution plan. I just kept the result in a variable for one minute, as the data don't change often. x1000 perf with 3 lines of code. I would have expected it to say that this query didn't have to be executed that many times in the first place. It codes well when you tell it what to do, but it is still dumb about what is the real problem (and so the real solution). That's worrying me a bit because Opus 5.5 clearly is SOTA, so better than any open weights models, and I really want to be able to run smart models locally on my laptop for independence and privacy reasons. So, if I still find Opus dumb now, when will I have an open weight model better than 5.5 that can run on my Ryzen 7 AI 350 with its Radeon 860m and 32GB of RAM? ... This summer with all the releases of DeepSeek and Qwen I was hopeful I could have it for Christmas, but it looks like I will have to wait much more...

1) Claude ($100) and ChatGPT ($20)

2) No.

3) I mainly use Claude as my main, Pi (with ChatGPT) for any automations. I manage them all with herdr. I cd into a path, kick off a claude/codex/pi session, then kill the session and manage it from anywhere with iceghost. (https://github.com/sharkjets/iceghost)

4) Negative.

5) I tried openclaw but found it a waste of my already scarce tokens. :D I build my own runners when I need a loop or cron-based automation.

1) I use 2x ChatGPT Pro 200, and a Claude Max 200.

2) I use other models via Open Router, but ChatGPT beats them in value (at the moment) thanks to Tibo's resets.

3) I mostly make Claude coordinate a group of Codex agents. They've set up a communication channel between them, and Claude acts as the coordinator and a second verifier. This saves me time because it's easier to talk to one agent than a group of agents. Codex parent session does only "comms" with Claude, and all the other agents run as sub agents.

4) For non coding usecases, I use https://www.ctlflow.com. This lets me run dozens of agents on tasks like equity analysis. The information a swarm of agents can surface from a large data store can be quite surprising.

5) I have tested OpenClaw, but now I just use ctlflow.com or claude + codex.