JB Bakst

Most software wants to read your email faster. JB Bakst wants it to know which emails you would have ignored anyway. His company, Chief, opens with a line that sounds less like a feature and more like a confession: your inbox is a twelve-year record of how you actually make decisions.

The product he is building takes that record and, in his words, plays it forward. Chief studies the way you have triaged, replied, ignored, and forwarded over a decade, then assembles a workspace each morning around what matters - not whoever happened to email you last. It drafts in your voice. It does the boring research before you open a thread. And it asks for permission in slices: forward the invoices, archive the promotions, leave the rest to you. The tagline is short and a little cheeky - the first inbox with good judgment.

Bakst is not a stranger to the unglamorous middle of productivity software. The reason Chief feels deliberate rather than trend-chasing is that he spent years on the inside of one of the tools that made spreadsheets feel like play.

The engineer who kept choosing delight

Bakst grew up in Manhattan and went west for computer science at Stanford, finishing in 2015. His first real job was the kind that teaches you everything because there is no one else to do it: a six-person startup called Chime, building a video-messaging app. He arrived to write iOS code and left having touched the full stack, design included. When a company is six people, the title on the door is just a suggestion.

In the summer of 2016 he joined Airtable as an engineer. The detail that gives him away is what he did before he started. He and his brother took a trip to Japan, and he planned the whole thing inside Airtable - the product he was about to help build. Most people audition for a job. Bakst auditioned the software.

At Airtable he gravitated to the seam between programming and design, the place where a database can either feel like homework or like a toy. He has described his interest plainly: building fun and delightful UX, especially for productivity tools. It is an unfashionable thing to care about. Spreadsheets are supposed to be serious. Bakst is one of the people who decided they did not have to be.

That instinct - that the dull software people use every day deserves taste - is the through-line. Email is the dullest software of all. It is also the one nobody has managed to make pleasant. So that is where he went next.

The word doing the heavy lifting there is trust. Plenty of tools will sort your mail. Far fewer earn the right to act inside it. Chief's whole design - permissions handed out in slices, drafts that sound like you, a curated workspace instead of a firehose - is an argument about trust. You do not hand the keys to an agent that treats every message as equally urgent. You hand them to one that has watched how you decide and learned to mimic your shrug.

There is a quiet ambition in reframing the inbox as a decision log rather than a to-do pile. It says the valuable thing in your email was never the messages. It was you - your taste, your priorities, the thousand small judgments you have made about what deserves a reply. Chief's bet is that this is the part worth automating, and the only part worth trusting.

What an inbox should weigh

Standard email ranks by one thing: whoever shouted last. Bakst's pitch flips the weighting toward judgment - the priorities buried in twelve years of how you have actually behaved. A rough sketch of the difference:

The kind of builder he is

Talk to the record and a shape appears. Bakst is the engineer who keeps wandering into the design studio, the full-stack generalist who would rather own a problem end to end than hand off the unglamorous half. He cares about the parts of software most people accept as ugly - the inbox, the spreadsheet, the daily grind - and treats them as places taste can win.

Chief is the natural endpoint of that temperament. It is not a flashy demo about what AI can do. It is a careful argument about what an agent should be allowed to do, and how it earns each new permission. For a founder, that restraint is its own kind of statement.

Why he started with judgment, not speed

Every other assistant in your life races to be quicker. Faster sorting, faster summaries, faster replies. Bakst aimed somewhere stranger. The promise stitched into Chief is not that it will move quickly - it is that it will be right about what to skip. That is a harder thing to sell and a harder thing to build, because it asks the software to have a point of view, and to borrow that point of view from you.

Look at how he describes the morning. Chief assembles a workspace curated around what matters, not whoever emailed last. Read that twice. The default behavior of every inbox ever shipped is the opposite: newest on top, loudest in front, the squeaky wheel rewarded with your attention. Bakst is proposing that recency is a bug we all agreed to call a feature. The fix is to rank by the priorities you have demonstrated over years, not the timestamps strangers control.

The second move is research. Before you even open a thread, Chief is supposed to have pulled the context - the calendar, the prior emails, the relevant material from your connected tools - so the question in front of you arrives with its homework already done. Anyone who has reconstructed a six-month-old conversation just to answer one line knows the tax this removes. It is unglamorous work. Bakst has made a career out of taking unglamorous work seriously.

Then the drafts. Not generic, helpful-assistant prose, but replies shaped to your communication style and the way you tend to decide. This is where the twelve-year decision log pays off. The same record that tells Chief what to surface also tells it how you would answer. The ambition is a draft you can send with a glance rather than a rewrite.

And finally, the part that separates a toy from a tool he would actually trust: permissions in slices. Forward the invoices automatically. Auto-archive the promotions. Leave everything else to a human. Bakst is explicit that this is not full autonomous control handed to a black box. It is a dial you turn one notch at a time, as the agent earns it. For a product built in 2026 - a year crowded with confident claims about agents doing everything - that restraint reads almost like a thesis statement. The fastest way to lose someone's inbox is to deserve it before you have earned it.

It is worth noting what Chief is not. It is not a productivity-theater dashboard. It is not a chatbot bolted onto a mail client. The pitch is quieter and more confident than that: most of what you do in email is busy work - triage, research, the same kinds of replies - and the genuinely valuable part is the judgment underneath it. Automate the busy work, surface the judgment, and the inbox stops being a place you survive and starts being a place that works for you.