Sathish Krishnan

Co-founder, Applied AI @ Tracfox

[Sathish Krishnan @SatyKrish]

The model isn’t the system. The coding agent isn’t the system. They are part of the solution. Your agentic workflows shouldn’t go down with them.

Recent Grok and Claude outages, along with OpenAI’s planned cutoff for Cursor, highlight why portability matters.

I have been building an agent platform where a harness coordinates a pool of ephemeral workers. Each worker can run a different coding agent and model combination: Claude | Codex | Grok | Muse.

Run state, traces, and artifacts are persisted in object storage across workers.

The harness owns the outer loop: scheduling, execution graphs, state management, retries, evaluation, reassignment, and multi-agent coordination.

The worker owns the inner loop: a coding agent capable of spawning its own subagents to complete the assigned task.

This came in handy recently when I switched from Grok to Muse Code without changing the rest of the workflow.

As builders, we should be designing for this from day one. Build the harness, own the workflow, and treat models and coding agents as replaceable parts.

[Sathish Krishnan @SatyKrish]

Stop benchmarking models in isolation. Benchmark the model + harness.

I’ve been testing GPT-5.6 Luna xhigh through Codex on a fairly complex agentic workload. The Codex harness uses separate Planner, Generator, and Evaluator agents.

For the Luna/xhigh run:

- Runtime: 9m42s
- 4 independent sessions: coordinator + - Planner + Generator + Evaluator
- 10 agent messages
- 9 collaboration calls
- 1.88M tokens processed
- 94%+ input tokens cached
- 14.7K output + reasoning tokens
- Total cost: $0.074

More importantly, it worked really well.

The model matters. But the harness determines how much of that model you actually get. With Context engineering, Caching, Tools, Evaluation and Retries, a good harness can make a smaller, cheaper model punch well above its weight.

Nice work, @OpenAI.

[Sathish Krishnan @SatyKrish]

@gdb ChatGPT plugins with MCP-UI and Skills over MCP makes it really powerful to launch your products on @OpenAI’s distribution.

[Sathish Krishnan @SatyKrish]

Agent skills have a distribution problem.

Today, harnesses often need skills to be packaged, installed, or pulled from a separate registry.

Skills Over MCP offers a cleaner model. It serve skills as MCP Resources and let the harness discover them dynamically.

ChatGPT Plugins already support this natively.

MCP as a distribution layer for agent skills is the right direction.

Ref: https://t.co/c0ObRlu3cw

[Sathish Krishnan @SatyKrish]

@TheConn_Man It’s backend to the product. Handling thounsands of document processing autonomously.

[Sathish Krishnan @SatyKrish]

Environment matters as much as the context we provide to the models. I have been building agentic system with harness at its core, managing a bunch of ephemeral workers. Technically they are containerised environment with a coding agent, tools (CLI/MCP), mounted paths (for corpus) and model.

Harness chooses worker appropriate for a task, and provides task specific skills and execution graph (for complex multi-agent flow). Self improvement is done by grading output and observing reasoning traces. Hooks govern tool permissions and handle tool result.

Able to run multi-turn long horizon tasks. Now looking into multi-day unattended runs.

#AgenticAI #HarnessEngineering

[Sathish Krishnan @SatyKrish]

Reviewing reasoning traces reveals a lot about how context, skill and instructions can influence coding agent performance .

So, before blaming a model for poor outcome, check what you are feeding it.

[Sathish Krishnan @SatyKrish]

@thedarioaltman @DanielSmidstrup I agree. Focus on the workflow, while using model and coding agents interchangeably.

During recent grok and Claude outage, I quickly swapped my workflow to run on Meta’s muse code.

[Sathish Krishnan @SatyKrish]

@dxiaolong Initial planning, mid-run feedbacks and auth. My codex mobile usage is about 20 hours across 5 days.

[Sathish Krishnan @SatyKrish]

OpenAI ’s Codex on ChatGPT app has changed how much I can get done from my phone, without sitting in front of a computer.

With Astra’s computer use capabilities, Codex manages end-to-end SDLC: building, testing, reviewing, and deploying.

Phone is becoming my daily driver. Agents handle the execution while I steer the work.

@OpenAIDevs how has been your experience?

My week’s usage stats:

[Sathish Krishnan @SatyKrish]

@dxiaolong Full SDLC, with Codex running on Mac Studio.

[Sathish Krishnan @SatyKrish]

@benln @bot 🙋‍♂️

[Sathish Krishnan @SatyKrish]

@TheConn_Man This is powering our Product, serving FIs.

[Sathish Krishnan @SatyKrish]

@TheConn_Man k8s is optional. Agents don’t follow the same scaling rules as pod. You can however make them stateless, requires thoughtful design.

In this case, workers are container instances powered by microVM. AWS Agentcore, Lambda MicroVM or applecontainer on Mac.

[Sathish Krishnan @SatyKrish]

@TheConn_Man Custom built for the workflows I manage.

[Sathish Krishnan @SatyKrish]

@paulg Outcome, distribution and easier deployment can make a huge difference.

[Sathish Krishnan @SatyKrish]

@lovejing0306 Coordinator uses model router.

[Sathish Krishnan @SatyKrish]

The model isn’t the system. The coding agent isn’t the system. They are part of the solution. Your agentic workflows shouldn’t go down with them.

Recent Grok and Claude outages, along with OpenAI’s planned cutoff for Cursor, highlight why portability matters.

I have been building an agent platform where a harness coordinates a pool of ephemeral workers. Each worker can run a different coding agent and model combination: Claude | Codex | Grok | Muse.

Run state, traces, and artifacts are persisted in object storage across workers.

The harness owns the outer loop: scheduling, execution graphs, state management, retries, evaluation, reassignment, and multi-agent coordination.

The worker owns the inner loop: a coding agent capable of spawning its own subagents to complete the assigned task.

This came in handy recently when I switched from Grok to Muse Code without changing the rest of the workflow.

As builders, we should be designing for this from day one. Build the harness, own the workflow, and treat models and coding agents as replaceable parts.

[Sathish Krishnan @SatyKrish]

There’s huge excitement about GPT-6 Astra’s capabilities.

Astra at low reasoning does the work I used GPT-5.6 Sol high or extra. @OpenAI‘s own evals show stronger results with fewer output tokens. I default Astra to low effort, and raise effort only when the eval says so.

Why does it matter?

Reasoning tokens are the model’s internal chain-of-thought computation, but they are not exposed as readable output. They are generated internally before visible output, and billed at the model’s output-token rate. Higher effort generates more reasoning output.

So if usage reports:

{
"output_tokens": 11860,
"output_tokens_details": {
"reasoning_tokens": 10524
}
}

you pay for all 11860 output tokens, including the 10524 hidden reasoning tokens.

At Astra’s pricing if $10 per 1M input tokens and $50 per 1M output tokens, higher effort can burn through your budget sooner.

#Tokenomics_101

Astra is stronger at general instruction following.

It is recommended to align AGENTS.md and Skills as per prompt guidelines before using Astra in your agents.

https://t.co/I4elr4RJ2e

[Sathish Krishnan @SatyKrish]

@JoshKaplanM Waiting …. 😮‍💨

[Sathish Krishnan @SatyKrish]

@Vyom_B @thsottiaux Podcasts while waiting on Codex.

[Sathish Krishnan @SatyKrish]

@mntruell @hnshah Grok Bot been dogfooding my apps. Can easily be App Support too.

Easy setup. Great value.