Posted by ssiddharth 3 hours ago
So much for AI getting cheaper.
For now, that's DeepSeek: https://api-docs.deepseek.com/quick_start/pricing/ (they have a discount until the end of the month, even after that they will have pretty good prices)
Or GLM or Kimi, Mistral is also surprisingly passable. Or just have to open the wallet and give money to OpenAI or Anthropic for the subsidized tokens.
> Google Enterprise accounts are sunsetting AI Ultra in favor of consumption based pricing at the end of the month.
This whole thing feels a bit like what GitHub did with Copilot, though.
There's your mistake right there. There is history. User beware.
It's centered around git worktrees. The goal is to organize all your AI sessions into somewhat logical places and make it easy to context switch. The secondary goal is to remove the need to open a separate code editor anytime you want to look at a file (We have a built in file editor powered by Monaco [vscodes editor])
Check it out https://harness.mikelyons.org
Isn't this what Pi does (except you have a non-CLI UI)?
But I have to say that I never understood the Antigravity IDE. I much prefer using Gemini CLI in combination with vscode. It works like a charm. Now, I'll do the same with Antigravity CLI and vscode. It works fine.
Cursor still supports both the IDE and the Agents window, open at the same time, in the same project. I frequently use both and switch back and forth between them. They also link to each other from the top bar and right-click context menus so you can switch to one or the other seamlessly. Best of both worlds. Switch back to Cursor.
Google could easily A/B test half of their users away from their products and nobody would get fired for it
This was clearly an experiment or stepping stone, they were never going to stick to this path. It was always going to go away.
The problem with AI products vs other rent-seeking is that AI is very expensive to build out and run… so they are desperate to push you into relying on it quickly.
Overall the experience was pretty bad for what is expected from them and I'm wondering what the thought process behind this is, I dislike this single prompt box review workflow and is a reason I don't use any of the tui stuff and it's odd that they are leaning so hard to mimic CC when others like cursor are embracing the same workflow but still sculpting around the code. I want to edit as I'm working and have access to all my normal tools and fragmenting my work to this new vision and a separate text editor defeats the point.
For now I'll probably switch to using it as a fallback when I've exhausted my quota elsewhere and start to rely on it less before the next rug pull when I wake up and the IDE is gone. Aside, Gemini has been surprisingly good and I really liked their take on the implementation and review workflow.
AI is powerful, but currently does not meet the engineering bar for quality and thoroughness. We need new paradigms and tools to support a new relationship with the codebase as an artifact.
The premise is that we can use these LLMs to get real engineering work done if we make tools to support a higher-level human understanding of the codebase, and the ability to spot the gaps in the LLM's plans. With these we can surgically ensure all the critical considerations are covered, spec the work at an incredibly granular level, and implement our plans as a collection of ultra-tiny tasks each given to isolated agents, this specifically ensures the agent's attentional mechanism aren't overwhelmed/polluted.
The project is very early still, so if you're interested, please reach out or signup for the email-list and i'll contact you. Pricing page is highly aspirational at the moment, money is not the focus at this phase.
Thanks.
Why do you think an IDE is the right tool?
I'm working in a similar space, and it's not clear why an IDE would benefit.
Specifically to you - if you're hoping to make this a business - please know if you do make a killer IDE feature - Cursor et al will immediately copy it...
I'll give your tool a try if it's not too much effort to try it and you want some feedback. Let me know.
I didn't start with an IDE but ended up there after some time. The core of my approach is an entirely new workflow. Underlying all of it is a "planning canvas" which is a network graph visualization of the codebase symbols, structures, and relations, where each node of the graph is a custom data-structure that captures a set of considerations. The workflow is generally as follows: Talk to the agent -> Agent responds with a plan(s) -> Plan is visualized on the planning canvas. At this point we can see visually which parts of the codebase the agents plan touches and via the fields of the custom data-structure, also see which considerations the agent failed to specify. Its here where we as humans can catch "this thing isnt connected, or is missing a trigger, or has a concurrency story, etc.", and either specify ourself, or force the agent to improve their plan in this specific manner. Once satisfied, we can formalize the impoved plan into a spec-of-specs, where each isolated sub-spec is farmed to an agent for implementation, which undo/redo being handled at the plan-level just in case we change our minds.
> Cursor et al will immediately copy it...
This is always possible, with anything and everything, but thus far they havent done it and i want this to exist, so i persist.
> I'll give your tool a try if it's not too much effort to try it
If you're open to it, signup (so i have your email) and ill reach out to get us going.
Cool, I'm thinking along the same lines.
> but thus far they havent done it and i want this to exist, so i persist.
Cool, we are in the same boat [=
> If you're open to it, signup
I'll check it out.