Zerobuild

The workspace with memory.
Team chat, customer conversations, and docs in one tool.

[Zerobuild @Zerobuildai]

@klashsarma https://t.co/bEt1qRbKG4

[Zerobuild @Zerobuildai]

@jack gets the problem exactly right with buzz: "agents can't help with what they can't see." context is the bottleneck, not the model.

buzz fixes it the way an engineering org would — host your own relay, sign every event, pull code, CI and chat into one place. great for block. but he says it himself: a hosted option for teams that don't want to run infrastructure is still ahead. that's most small businesses. same problem, right now, and no relay to run it on.

we're coming at it from the other side. not the eng stack, the front office - customer conversations, the internal thread, the doc, all in one place because that's where the work already happens. your AI can read it today. nothing to host.

same diagnosis. you just shouldn't need an infra team to own your own context.

[Zerobuild @Zerobuildai]

@jack gets the problem exactly right with buzz: "agents can't help with what they can't see." context is the bottleneck, not the model.

buzz fixes it the way an engineering org would — host your own relay, sign every event, pull code, CI and chat into one place. great for block. but he says it himself: a hosted option for teams that don't want to run infrastructure is still ahead. that's most small businesses. same problem, right now, and no relay to run it on.

we're coming at it from the other side. not the eng stack, the front office - customer conversations, the internal thread, the doc, all in one place because that's where the work already happens. your AI can read it today. nothing to host.

same diagnosis. you just shouldn't need an infra team to own your own context.

[jack @jack]

https://t.co/WFWSz5jEpc

[Zerobuild @Zerobuildai]

@bensenescu @jack We are taking a shot on bunch of them including slack.

[Zerobuild @Zerobuildai]

@jack We felt the same and ai era needed a new unified workspace.

We are building one

https://t.co/rdOboPYigT

[Zerobuild @Zerobuildai]

https://t.co/KNagh3Ld1F

[Zerobuild @Zerobuildai]

@amasad re-bundling in action. ai works way better when it doesn't have to borrow context from external services.

One belief zerobuild is built on, ai makes software cheap enough that you can build a lot of it yourself. a unified access for ai.

[Zerobuild @Zerobuildai]

The self-driving company is real.

But Replit needed an agent harness, microVMs, ZeroTrust, token proxies, and eight integrations to get there - a platform team's output.

A 15-person shop can't build that. If the work already lives in one substrate, it doesn't have to. That's the bet.

https://t.co/vfvdqSFOJr

[Amjad Masad @amasad]

https://t.co/9ujCmLs3lQ

[Zerobuild @Zerobuildai]

Clearest version of the moat argument I've read; private ground truth, not the weights.

The mechanism you describe (get inside, do the unglamorous translation, engineers next to the customer) is enterprise-shaped. A 5-person company can't buy that and doesn't have to: when the workspace is the company's private reality, the translation is structural, not a services engagement.

Argued the small-team version of this last week. You don't integrate your way to the untrainable — you own the substrate it lives in.

https://t.co/vfvdqSFOJr

[Zerobuild @Zerobuildai]

https://t.co/KNagh3Ld1F

[Zerobuild @Zerobuildai]

@paulg @naval Yes we’re doing exactly that.

[Zerobuild @Zerobuildai]

https://t.co/KNagh3Ld1F

[Zerobuild @Zerobuildai]

A fine-tuned vertical model bets the gap is capability.

But frontier models get more capable every quarter — what they lack is your context, not reasoning. The model freezes at training time. Your workspace compounds every day.

Bet on the model, you bet against the frontier.

[Zerobuild @Zerobuildai]

https://t.co/KNagh3Ld1F

[Zerobuild @Zerobuildai]

Two ways to make AI expert at your business:

Put the expertise in the model (fine-tune a vertical model).

Or put it in your workspace (let a generalist read what your company wrote down).

Same goal. Completely different layer. Only one of them is yours.

[Eoghan McCabe @eoghan]

https://t.co/I2Hf40EpIt

[Zerobuild @Zerobuildai]

https://t.co/KNagh3Ld1F

[Zerobuild @Zerobuildai]

Yes and worth naming the mechanism. Domain knowledge lives in people's heads because the software we work in throws the context away as we go.

You don't have to extract it. You have to stop discarding it.

@paulg's right that AI-native companies skip the extraction step because their work lands somewhere a model can read, from day one.

[Tom Blomfield @t_blom]

Imagine replacing 90% of your employees with a team of geniuses who have no idea how your company operates.

Total chaos. Nothing works.

That’s what AI feels like today.

The missing piece is extracting all the domain knowledge from people’s heads and providing that as structured context to the models.

[Zerobuild @Zerobuildai]

The companies that build for this open intelligence as the default, on a shared workspace where humans and AI both read and write, won't need to hire fifty people to do what fifty used to do.
Team gets smarter. AI gets verticalized. Same workspace, at the same time.

[Zerobuild @Zerobuildai]

https://t.co/5JBfy8Nnqx

[Zerobuild @Zerobuildai]

Here's the full article from Tobi

https://t.co/BQvUch9phc

[tobi lutke @tobi]

https://t.co/5MJ7u1VHwf

[Zerobuild @Zerobuildai]

Spent the weekend on this one.

@tobi Lütke shared a number from Shopify last week that's been stuck in my head — 36% to 77% accuracy on River's merge decisions in two months, no model change.

It points to an architectural fork the next five years of work software run through.

[Zerobuild @Zerobuildai]

https://t.co/5JBfy8Nnqx

[Zerobuild @Zerobuildai]

https://t.co/5JBfy8Nnqx

[Zerobuild @Zerobuildai]

Read more at our blog -> https://t.co/RI8xAHkLqg

[Zerobuild @Zerobuildai]

https://t.co/A3WXCzhpIR