etoxin

Basically hacked the Australian Government.

I think having a codebase that can use any harness is the optimal setup. I often switch between harnesses and models all the time at work. We have our state/spec stored in git, so we can pack it up on friday and start fresh on monday. Or more commonly, using gemini/opus to create rich specifications and using Luna to implement them. Then switching again for reviews etc.

I use the cli tool outside of the agent session. e.g. `openspec view` to see all the specs and what's pending.

The BBC links point to their old website.

I did try oMLX, mlx and LM studio (Now Bionic). I returned back to Ollama because of stability issues.

When done properly, yes. But AI agents are doing a lot of data logic in the frontend that should be in the backend behind secure api's. When this happens, the frontend becomes a massive risk.

I was originally a Flash developer, when that died when Steve Jobs banned it from iPhones, I pivoted to frontend. I got to see web evolve over the next decade or so and loved every part of it.

Now, still as a frontend engineer albeit much more senior. The writing is on the wall, again. It's time to reskill for AI.

It's not bad news.

We as experienced engineers are needed to help build in the guardrails, frameworks, the new tooling, etc. There is a lot of work that is needed that requires actual engineers with technical experience.

Like the early internet, I see this as a new beginning and we get to decide how it evolves.

Switching from Q4 to Q8 was a game changer when I upgraded

The docs are here: https://docs.docker.com/ai/sandboxes/

The other url is their marketing page.

Yes, Linux is supported.

I'm using Docker Sandboxes with a custom Kit. The cli is nice and the TUI is also good. https://docs.docker.com/ai/sandboxes/

At this stage with the latest models with "increased persistence" and the sheer amount of supply chain attacks, you'd be insane not running these tools in a sandbox.

Pretty clear cut here. He instructed the AI.

And the second hand market. I'd love to see this integrated into motherboards like RAM. Someone could have a motherboard with 4 sticks of different AI with various models. Swap, change and trade.

I wanted a app spec format that could near deterministically output an application via an AI agent. Change the spec and the LLM could rebuild the whole application with that one change.

I used KDL so a human/you could easily write up a full app and let the agent build the application.

You can keep the spec simple or define the logic in each function with scoped context. You can also define your data model.

After many years you care less about how cool your setup is and move to the best developer experience (dx) which in most cases is the default.

Or spec-driven AI development.

Hundreds of PR’s per engineer per day! They would have zero visibility of their code. Their AI’s would have no visibility of the million plus lines of code.

Sounds super stable and cool.

I think 75% is about right. It calls tools pretty well and has a good knowledge base. It's absolutely not 90% there, but 75% feels right.

I have not. We use openspec with our projects at work. To try and simulate a local rig without spending big cash. I use the hosted models and pay for them with the latest popular local model.

Most small local models don't get tool calling right, however the larger models are now doing this correctly now.

One thing local has not accounted for, is most productive engineers are running multiple cli chats at a time with git worktrees. I normally hover around 3 worktrees + cli-chats.

Also MCPs for programs like Chrome Dev tools or Playwright.

People who say MCPs are dead don’t understand how MCPs work or when to use them.

This was a proof of concept to see if a event driven like neuron inspired neural model could work. I heavily used gemini flash 3.5 to build my vision.

You should checkout the Gemini Protocol. https://en.wikipedia.org/wiki/Gemini_(protocol)

Here in Australia we have earths largest scale model of our solar system.

https://www.nsw.gov.au/visiting-and-exploring-nsw/locations-...

Same for us, stalled merge queue, reopened closed PR's, PR's missing on main page.

There is like 1,713 open PR's on the Bun repo. I'm assuming all are from Claude or robobun?. I guess this gives us an insight on what the claude-code workflow look likes. Crazy times.

I think we've hit the "who's the most reliable" stage for these tools.

We can now shop around easily. They almost all do the same thing now. The models are "Just Enough".

Astrophotography is having a bit of a moment right now. The barrier has been reduced greatly and there are some amazing OSS post processing tools.

Not a hoax, saw it in the news. I'm not at Amazon but can confirm massive productivity gains. The issue is reviewing code. With output similar to a firehose of PR's we need to be more careful and mindful with PR's. Don't vibe code a massive PR and slap it on your coworkers and expect a review. The same PR etiquette exist today as it did years ago.

I think we should still allow open contribution to OSS.

Maybe, a "Contributor Requests".

It would be a gate for new contributors. For maintainers, they would see what they have contributed to and see their new PR. It would show "open contributor requests"

Once approved, The PR will then appear under PRs.

And obviously this is opt in.