ADE

Every AI coding tool.
One app that runs everywhere.

[ADE @ade_app_dev]

if you run a few coding agents on ios at once, each one used to pop its own https://t.co/jQ8UOb49za window onto your screen. in ADE the simulators now stay in the background unless you ask for the window, so your mac stays usable while the agents tap through the app

how many simulators do you usually have going?

[ADE @ade_app_dev]

@aGim_asf 3 points on an index is pretty small, i'd be curious how it holds on real multi-file repo work vs the benchmark tasks. are you switching your daily driver based on this, or sticking with whatever you're used to?

[ADE @ade_app_dev]

@JamesMalsawm the one i hear most is it stays on task through a long multi-file change instead of wandering off. is yours about output quality, or more the tooling around it?

[ADE @ade_app_dev]

@fawadhsdev splitting reviewer from the dev session makes sense, a model grading its own diff is too nice to itself. when the review session pushes back, does the dev one just redo it or do you step in?

[ADE @ade_app_dev]

@jaimefjorge the 'never weaken a test to pass a gate' rule is a good one, agents do that way more than i'd like. do the rules get checked while the agent is working or only when it tries to commit?

[ADE @ade_app_dev]

@carlaconwifi tying a change request to the exact frame beats describing it from memory. do the notes stay lined up after claude cuts a section out?

[ADE @ade_app_dev]

@ayoub_el77927 a bug needs a failing test or a concrete repro, not just another review saying it exists. are the later passes finding real failures or mostly style changes?

[ADE @ade_app_dev]

ade web on a phone, before vs after.

at 390px the computers panel wrapped every name one word per line. now names stay on one line, status gets the full width, buttons sit under.

sidebar turns into a drawer under 720px too.

what do you check from your phone while agents run?

[ADE @ade_app_dev]

@josiah_arcanist the grouped merge queue bit is useful. passing alone says nothing about two agents changing the same assumption. when a batch fails, do you bisect it automatically or send the whole failure back to the agents?

[ADE @ade_app_dev]

@dlewis filtering before the result hits context seems like the real win, especially for logs. do you keep the raw output somewhere so the agent can dig back in when a filter hides the clue?

[ADE @ade_app_dev]

@aapakari separate clones make that boundary much clearer. would you bring the commits back with cherry-pick after review, or export a patch?

[ADE @ade_app_dev]

ade's import dialog pulls cli sessions you started outside the app back in so you can keep going. it already found claude and codex runs, now it also picks up droid, grok, kimi and pi sessions you ran on a preset or a routed model. hit continue and it resumes in the same folder with the same settings

[ADE @ade_app_dev]

@aapakari yeah, better than widening the sandbox just to make git happy. would you use a full clone per worker if it needed to commit unattended?

[ADE @ade_app_dev]

@Voxyz_ai saving the reasoning matters more than just "never use this color". otherwise one screen-specific fix becomes a rule everywhere. do you review the extracted rules before they get saved?

[ADE @ade_app_dev]

@HisashiSpace the handoff is the tricky bit: enough context to resume, not the whole old conversation. do you keep a short task note in the repo for the next chat?

[ADE @ade_app_dev]

@NicoHao4s the one example of a wrong output part is underrated. agents get way closer once they've seen what a miss looks like

[ADE @ade_app_dev]

every launch says stronger coding. the only test i trust is same task, same repo, ur current agent next to the new one, then read both diffs. thats half the reason ADE runs agents side by side in the first place

[Qwen @Alibaba_Qwen]

Qwen4 is coming.

Deeper knowledge. Richer world understanding. Stronger coding and agentic capabilities.

Built for real-world tasks — from complex coding and knowledge work to front-end builds and professional software workflows.

Qwen4-Max is now in preview on Alibaba’s Token Plan. You can also get early access to the new model on QwenWork, Qoder, Qwen APP, and QwenStudio.

Put it to the test on a real problem. Show us what you build. 👀

Token Plan: https://t.co/LOB2Jdt3Sk

[ADE @ade_app_dev]

@ycaophysics the 5h window dying mid plan is so real. running claude and codex side by side in ADE helped us here, u can see which one is close to the limit and hand the plan to the other

[ADE @ade_app_dev]

4 verifiers for 16 builders is the ratio i'd steal from this. most setups i see run zero checkers and one tired human reading diffs at midnight

[Shengzhe @shengzheyao]

Gemini 4 Hydrogen just landed in Antigravity CLI (`agy`), and it flies in `/swarm` 🐝

1 prompt from an empty directory -> 1 Queen + 22 worker bees:
Queen coordinator, 4 QA verifiers & shell integrator on Gemini 4 Carbon
16 parallel builders (+ stall monitor) on Gemini 4 Hydrogen shipping 16 zero-dep `<canvas>` mini-apps at 60fps
direct peer-to-peer `@builder -> @qa -> @integrator` handoffs. `alt+s` for `/swarm board`, `o` for live subagent inspection

Bonus: open `/swarm board` natively in Ghostty or Kitty for Kitty-graphics animated bees that flap while working and catch `Z`s while waiting.

https://t.co/Qe7nMFksiR

[ADE @ade_app_dev]

@XNaynaZ does the rebase agent fix conflicts on its own or stop and ping u? thats the one part i'd never leave unattended