GPTree

[GPTree @GPTreeAI]

GPTree has a new look.

New mark, new wordmark, new site, same job: branch a conversation to explore without losing your thread, compare models side by side, and merge the best parts into one answer.

Never re-explain yourself: https://t.co/vlChZG9vIV

[GPTree @GPTreeAI]

@kushmergedeck GPTree https://t.co/vlChZG9vIV

[GPTree @GPTreeAI]

@Seanwvibedev https://t.co/c4EAe7Z1Ko list for free - dogfood build by GPTree Graft + Foremerge

[GPTree @GPTreeAI]

@Seanwvibedev https://t.co/L51NEZV50F

[GPTree @GPTreeAI]

@patilprashhw https://t.co/1F4AWfQJ55

[GPTree @GPTreeAI]

@FlippedRay GPTree https://t.co/vlChZG9vIV

[GPTree @GPTreeAI]

@machoman385 Yes

[GPTree @GPTreeAI]

Automatic memory you cannot inspect is creepy, so GPTree built the opposite.

It keeps facts, decisions, patterns and identity from your conversations, each one scoped to a single conversation, a project, or your whole account. Everything stored can be seen, edited, pinned, disabled or deleted in the Memory Inspector.

Identity is the part worth knowing. Say something new about the same part of who you are and the old version is retired rather than kept alongside, so a job title you have moved on from stops shaping answers. Retired memories are not deleted. They stay marked as contradicted, so you can see what GPTree used to think and when it changed its mind.

Only your own words count. A description GPTree writes about you in an answer never becomes an identity memory, however confidently it is phrased.

https://t.co/gG1mxF8QEB

[GPTree @GPTreeAI]

@machoman385 Yes we always listened and try to absorb as much feedback as possible from our users

[GPTree @GPTreeAI]

Bring your own API keys and GPTree becomes the layer on top. Your key, your provider's bill, our branches.

Add OpenAI, Anthropic, Google AI, Mistral or Azure OpenAI on the AI Providers page, verify the key, and those models appear in the picker like any others. GitHub Copilot and Microsoft Copilot connect over OAuth instead of a pasted key. Keys are stored encrypted.

It pays off most in Explore, where one prompt runs as 2 to 5 side by side variants and each variant gets its own model. Comparing two providers on the same code review stops being a thought experiment.

Pro and Team, with a 14 day trial on Pro.

https://t.co/4WtUyRaBGr

[GPTree @GPTreeAI]

You are using browser tabs as a filing system for work you cannot afford to lose.

Not the conversations themselves. Those are saved somewhere. What lives only in the tabs is the map: which one had the good opening, which one you already abandoned, which one went stale three replies ago. Close them and the conversations survive, but you no longer know which is which.

Branching keeps the map on screen. A branch starts at the message it split from, carries the context behind that message, and sits in the tree beside the path you did not take.

The tabs were never holding the work. They were holding the order you understood it in.

[GPTree @GPTreeAI]

@JustinasRoland0 @X Try Foremerge if your building this weekend I promise you’ll thank me
https://t.co/L51NEZV50F open source 30s setup

[GPTree @GPTreeAI]

@Seanwvibedev https://t.co/L51NEZV50F open source

[GPTree @GPTreeAI]

@machoman385 Love to hear them

[Nick Woodhead | GPTree | Foremerge | TranslateTech @naw103]

I stop experimenting on a piece of writing at exactly the point it starts working, which is the worst possible moment to stop.

The reason is not mysterious. Once the draft is decent, the next idea costs something. It has to be tried in the same thread the decent version lives in, so testing it means agreeing to lose what I have if it does not land. I usually decide the current version is fine, and I am not sure I have ever been right about that.

Branching removes the wager. The working draft stays where it is, untouched, and the experiment runs beside it from the same message with the same context behind it. If the new idea is worse, nothing happened. If it is better, it was sitting one branch away the whole time.

The ideas I lost were never lost to a bad experiment. They were lost to not running one.

[GPTree @GPTreeAI]

@patilprashhw https://t.co/9plcxsG7Vn

[GPTree @GPTreeAI]

@patilprashhw https://t.co/9plcxsG7Vn

[GPTree @GPTreeAI]

@patilprashhw https://t.co/KVc9d7EVKs

[GPTree @GPTreeAI]

@Sarakhan49309 https://t.co/KVc9d7EVKs

[GPTree @GPTreeAI]

@JustinasRoland0 https://t.co/yNLIl91ntD

[GPTree @GPTreeAI]

@VitaliiRel https://t.co/yNLIl91ntD

[GPTree @GPTreeAI]

@ryanships https://t.co/KVc9d7EVKs