Dmitry Petrov

🛠️ Building data infra for AI/ML. Ex-Data Scientist @Microsoft. Created DVC, now DataChain. PhD in CS. Serious about data. Less serious about everything else.

[Dmitry Petrov @FullStackML]

@echpochmac8 Math with trits -1, 0, 1 is cleaner.

Physics becomes a mess.

My academic grand-supervisor worked on this. The fact that he moved on says a lot.

[Dmitry Petrov @FullStackML]

@toddsaunders https://t.co/JppovFHDjP

[Dmitry Petrov @FullStackML]

Opus 4.8 is really smart!

It spent 38 minutes implementing a pretty complex feature, proved it worked, then reverted it with a detailed explanation.

So, like humans 🤷‍♂️

[Dmitry Petrov @FullStackML]

@shikhr_ why not Opus 4.o-thinking-turbo?

[Dmitry Petrov @FullStackML]

Opus 4.8 is really smart!

It spent 38 minutes implementing a pretty complex feature, proved it worked, then reverted it with a detailed explanation.

So, like humans 🤷‍♂️

[Dmitry Petrov @FullStackML]

@TimJayas Another conspiracy theory got proven

[Dmitry Petrov @FullStackML]

@Nick_Davidov 🤔 I’m starting to understand why people dream about owning a horse

[Dmitry Petrov @FullStackML]

@Nick_Davidov That’s lane assistant, not autopilot

[Dmitry Petrov @FullStackML]

@dadiomov ok, staying at home today

[Dmitry Petrov @FullStackML]

@HerodotusWave Hello 911. I need to report a crime..

[Dmitry Petrov @FullStackML]

3am thought: https://t.co/S4fIw72c4P tonight. Pretty solid, and you pick your model GPT/Opus/DeepSeek.

The harness category is maturing into "bring your own model" mode. The harness IS the product now.

[Dmitry Petrov @FullStackML]

@wei_tianxin I’ll share the paper with more details once it’s ready

[Dmitry Petrov @FullStackML]

@bettercallsalva 💯 it already happened with code!

[Dmitry Petrov @FullStackML]

Every dev-experience problem ends-up generating and executing some code, and all code is moving under a harness (Claude Code, Cursor, Codex), leaving people with text.

Data platforms are next! And that's what we need to build.

[Tianxin Wei @wei_tianxin]

🚀Code as Agent Harness: A survey work from UIUC, Stanford, and Meta.

📄https://t.co/YReL1BMIoN

Code is no longer just the output of AI.

It is becoming the executable, inspectable, and stateful substrate through which AI agents reason, act, verify, remember, and self-correct over long horizons.

In our new survey, we examine this shift through the lens of Code as Agent Harness, focusing on how code serves as:

• 🧠 Harness Interface: coding for reasoning, acting, and environment modeling
• ⚙️ Harness Mechanisms: planning, memory, tool use, feedback, and optimization
• 🤝 Multi-Agent Harnesses: collaboration through shared code, tests, and execution traces

We review applications spanning:
💻 Coding Agents
🖥️ GUI/OS Agents
🤖 Embodied Agents
🔬 Scientific Discovery
🏢 Enterprise Workflows

If you find this survey helpful, feel free to explore our resource collection below.

🤗 Hugging Face Daily: https://t.co/cfuoQfzcj3
💻 GitHub: https://t.co/ZD156rWPbJ
🌍 Website: https://t.co/OfibqKL1en

Feedback, suggestions, and community contributions are warmly welcome!

#AI #Agents #LLM #Coding #AgenticAI #SoftwareEngineering

[Dmitry Petrov @FullStackML]

@walls_jason1 Yes, this is the whole point.

[Dmitry Petrov @FullStackML]

@omarsar0 Same four properties on the data side: executable (Compute Engine over data), inspectable (data UI), stateful (centrialized DB), governed (schema, version, lineage).

Data work needs its own harness. Measured 2.7x lower cost-of-failure under Claude Code with data context.

[Dmitry Petrov @FullStackML]

My productivity KPI: hit Claude's weekly limit. Anthropic doubled it. Now I'm underperforming 🫠

[Dmitry Petrov @FullStackML]

BigTech!

[Dmitry Petrov @FullStackML]

OpenAI's Data Agent and the S3 Gap https://t.co/d3oFALb8xO