Pre-seed · Open beta · Updated September 2026
Hadef is guest research for hosts of interview-style B2B podcasts. A host submits a name. Hadef reads every public appearance that person has made, finds the conversation nobody has had with them, and returns a pitch built on a verified verbatim quote from a real appearance.
This page is the short version of why that is a business rather than a feature, including the part we think is hardest.
What we are solving, and why.
Hosts of interview-style B2B podcasts compete for a small number of senior guests, and the quality of those guests is what the show is measured on. The obstacle is not effort, because hosts send plenty of messages. The obstacle is that the message which converts a director does nothing at the tier above, and the two failures look identical from the sender's side, so the response is to send more.
Our founder pitched at both tiers. The same style of message that booked managers and directors did not land a single guest at the senior tier he was aiming for, and working out why is what produced the method this product automates.
You get one attempt at someone senior. A weak approach does not cost you a reply, it costs you the person.
Volume tactics are rational when a name is replaceable. When a host has a list of twenty people who would change what the show is, and one attempt at each, volume stops being a strategy and starts being how the list gets burned.
On the numbers behind the senior-tier result: the sample is small and stated as a shape rather than a rate on this page, because a percentage drawn from a handful of approaches implies a precision the sample does not carry. The underlying counts are available on request.
How people do this now, and what we propose instead.
This is now the default. The advice that comes back is sound: build a target list, find warm paths, run a touch sequence, reference their previous appearances. Correct advice handed to every host at once becomes the floor, and the guest receives all of it. The single instruction that would separate one pitch from the rest, read what they have said and find the thread nobody pulled, is the one instruction a model cannot execute, because it cannot reach the audio and has no opinion about the show doing the asking. We published the full run of that experiment, prompt and output included, on the slop page.
It works, and it costs an afternoon per guest. Listen to the appearances, find the live thread or the buried one, verify the quote, work out what changed recently, then write questions the guest can answer without exposure. A list of thirty names, at an afternoon each, is a month of work, which is why almost nobody does it and why the hosts who do it are the ones booking the guests.
Automate the reading and leave the deciding with the host. The research is the expensive part and the part a machine can now genuinely do. The judgement about whether this angle suits this guest is the part hosts are good at and want to keep. Ninety seconds of research against an afternoon is the entire argument, and it only works because the output carries a quote the host can check against its source.
The claim everything rests on.
Everything else in podcasting is cosmetic by comparison. The questions are part of the pitch, so they are decided before the recording. Production quality is table stakes and has been for years. Listener growth follows guest quality rather than leading it. The higher the calibre of guest, the better a given show can become at being the thing it is.
A host who accepts that has already accepted the price, because every other lever they could pull is smaller. A host who rejects it will not be persuaded by a better feature list, which is why the claim sits early on every surface we write.
Hadef exists because the one lever that moves a show is also the one that does not scale by effort. Doing it properly is bounded by reading time, and reading time is the constraint we removed.
What Hadef does.
A signed-in host submits a prospective guest. Hadef finds that person's public appearances, transcribes them, reads them as a body of work rather than as search results, and identifies the conversation that has not been had with them yet. It then writes the approach in the host's own voice.
Those seven elements are the acceptance test for a finished pitch rather than a house style. A run that cannot produce them has not found an angle, and the host is told that instead of being handed something that fills the space.
Publishing the method in full is deliberate. A host who follows it by hand and books a guest without paying has proved the argument the product is sold on, and the subscription is then a question about time rather than about capability.
Why this is unclaimed.
Every tool in this space helps a prospective guest pitch podcasts. That is a real market and those products work, because the user wants the outcome and will do the pitching. The guest a B2B host actually wants is not trying to get onto other people's shows. They get approached, and they weigh whether the hour is worth spending.
What the category built
Directories of shows. Matchmaking by topic. Templates for the outbound message. Volume, because the sender's cost of a no is close to zero.
The user is motivated, self-selecting, and pitching is already their intent.
What a host actually needs
Research on one named individual who has not asked to hear from anyone. An angle nobody has used. A reason it is arriving today.
The target is unmotivated, and the sender gets one attempt, so volume is the wrong instrument.
This gap is invisible from where the rest of the category is standing. It only appears once you are the one doing the approaching and discover that every available tool assumes the other party wants to be found.
Saying your inbox is full of generated pitches comes close to saying your show is an easy target, so hosts absorb it silently and let the booking results speak. Senior guests have the same reason to stay quiet. The result is a problem nearly everyone in the category experiences and almost nobody has written down, which means the vocabulary for it is unowned and we can own it. The catalogue of named tells is the first instalment of that.
Who we are.
Hadef was built by Tan Sukhera, who spent the working span of a career on the craft of the approach and then applied it to booking senior guests for an interview show. The method on this site is the one he ran, and the workflow the product automates is the workflow he performed by hand.
The seventeen companies named in the repository's vouched-logo list, and shown as logos on the homepage, are guests that method booked by hand before any software existed. They are the evidence base for the product rather than decoration on it, and every one of them was reached by a human doing the reading.
Seventeen is the count of distinct companies in the repository's vouched-logo list, which is the same set the homepage carousel renders, and every one was booked by hand with this method. The carousel repeats the set to loop, so counting marks on screen gives thirty-four. It is not a customer count and is not presented as one.
The team, stated plainly.
Hadef is Tan Sukhera. There is no second founder and no engineering team, and we would rather you read that here than discover it later. The product exists at its current scope because the build itself is heavily automated, which is the same argument the product makes to its customers, applied to the company that makes it.
What that buys is speed and a very low burn. What it costs is the resilience a second person provides, and it is the first thing a raise would address. We have a view on the order of those hires and it is a conversation rather than a paragraph.
Why we are best suited to solve this.
Most products in this shape begin with a hypothesis about a workflow. This one began with the workflow, executed manually and repeatedly, at a close rate that made booking senior guests routine rather than exceptional. The logos exist and the episodes are public. What Hadef automates is a sequence that was already producing the outcome, which removes the largest question a product like this usually carries.
The open question is whether other hosts will pay for the automated version, not whether the method works.
That is a narrower risk than most pre-seed software carries, and it is also a real one. We would rather be judged on it directly than have it folded into a general claim about product-market fit. What we can say is that the failure modes we would be debugging are commercial ones, and the technical and methodological questions are behind us.
What we think is hardest.
The reason nobody has written about this problem is the same reason nobody is searching for a solution to it. There is no expressed demand to capture, because expressing the demand is self-incriminating. That makes the work ahead education rather than capture, and education is measured in quarters, costs more per customer, and cannot be short-circuited by outspending a competitor on the same keywords.
Any investor would reach this conclusion within a week, so naming it first is worth more than the risk it carries. The rest of the list is on the same basis.
Largest
The content strategy is the go-to-market rather than support for it. If the vocabulary does not take, the funnel has no top, and that is the failure mode we watch most closely.
Transfer
The method worked in one operator's hands on one show. Whether it holds across hosts with different subject areas and different standards for a good guest is the thing the beta is built to find out.
Concentration
Stated above and repeated here because it belongs on a risk list. It caps how fast the company can absorb a good outcome as much as it caps the bad ones.
Dependency
The pipeline reads public material through third-party services. Their pricing and their access terms are inputs we do not control, and the unit economics move when they do.
Status.
The product is live at hadef.io. Hosts apply, applications are reviewed within a few days, and small cohorts are admitted every few weeks. The cohort structure exists so that each intake can be watched properly rather than to create scarcity.
We are preparing a pre-seed raise. The use of funds is the content engine that creates the demand described above, and the first hires that remove the concentration risk named above it.
The material behind every claim on this page, including the counts, the close rate on the manual method and the current beta cohort data, is available to anyone having this conversation with us.