Justin Tang

cto & co-founder of @LanternAI__ , backed by @YCombinator (F26)

[Justin Tang @JustinTang29]

building something you’re proud of takes the right people too. lantern.md helps teams find and reach out to candidates who could be a great fit.

[Justin Tang @JustinTang29]

sam altman admitted he makes this founder mistake at @ycombinator dinner talk

he often knew when a project needed to end but kept finding reasons to delay the decision.

even with the answer in front of you, “let’s give it another week” feels easier than pulling the plug.

easy to understand why, esp since changing direction means admitting that some of that work won’t lead where you hoped.

sam said part of YC’s value is having partners and other founders who push you to make those decisions sooner..

...and that is harder to recreate once the batch ends.

his advice at YC: "get better at doing the hard thing sooner"

making that call also creates room for the next opportunity, and that’s a mindset I want us to carry as we build Lantern (yc f26)

we might not get every decision right but being willing to change our minds gives us a better chance of building something we’re proud of.

[Justin Tang @JustinTang29]

Y Combinator F26 welcome party ✅

Before the batch started, I’d see what other founders were building and think: damn, everybody moves fast.

Everyone’s figuring stuff out as they go, but nobody’s sitting still. Being around that kind of energy definitely makes you want to raise your own bar.

Big thanks to @insforge and @sprintoHQ for hosting this with us, and everyone who came!

back to building @LanternAI__ 🔨

[Thomas Chan @onethomaschan]

robert pattinson has primetime coming out, so does Lantern AI (YC F26)

except ours is about an interrogation over why a role is still open.

today we're launching Lantern agentic sourcer.

hiring managers know who they want:

> a scrappy ex-founder with product ownership
> gtm engineer who's built scalable outbound
> someone who can turn a rough idea to something working overnight

but it’s next to impossible to express this in filters and boolean searches.

Lantern makes your hiring taste searchable.

It learns your taste, looks across the talent market to track down people you’d actually hire, then reaches out to them.

naturally we decided to make a primetime inspired launch vid about it ;)

hire with Lantern: www.lantern.md

[Thomas Chan @onethomaschan]

this person injected a prompt in their LinkedIn bio asking recruiters for a flan recipe.. and it worked lol

but deleting the recipe still wouldn’t make it good outreach.

recruiters/ hiring managers need to know:

> what have they done that’s relevant for this role?
> what am I assuming about them that might be false?
> why would this opp be worth considering from THEIR side?

We ask candidates to explain why they want the job. We should be able to explain why we contacted them.

[Justin Tang @JustinTang29]

this is a good example of why we need better market intelligence when hiring.

Say you’re hiring an ai engineer. before opening that role, I wanna know:

- which companies have the best people doing this today?
- what are they paying?
- is my comp range realistic?

you can answer most of these questions yourself. but it's gonna take a lot of research.

we're building the easiest way to do this. before you start recruiting, lantern.md looks at the market for the person you’re trying to hire.

maybe your comp might be too low. maybe you're searching for the wrong title. the person you’re describing might not even exist.

better to find that out before you start recruiting

https://t.co/fG2QGe2ErF

[Samuel McDonnell @samueljmcd]

Founders confuse three roles, not two:

1. ML researchers (train models)
2. AI engineers (build systems around models: evals, RAG, agents)
3. Product engineers with AI tools

Most startups hire #1, need #3, and don’t know #2 exists. The middle one is where projects live or die.

[Thomas Chan @onethomaschan]

Since joining @ycombinator, we've received lots of interest from investors before the batch even started.

One piece of advice from YC resonated a lot with me: fundraising and valuation are not measures of startup success.

The real work is:
- Talking to customers
- Building the product
- Creating sustained momentum

@JustinTang29 and I are grateful and honored, but our priority rn is building @LanternAI__

Capital should help a great company move faster. It shouldn’t distract you from building one.

[Justin Tang @JustinTang29]

"half of you in this room have a serious mental health condition."

@bryan_johnson was probably right

the ai industry's favorite productivity model is simple:

sleep less → work more → feel worse → hide it → call it grinding

this doesn't produce resilient teams. it produces people who can ship but is losing the capacity to switch off and recover

go hard, but not at your own expense.

p.s. he's 49 with a biological age of 28 🤯

[Justin Tang @JustinTang29]

catchup with fellow hk native yc founder. cto of bricksAI (yc s22) and founding engineer at fintool (yc w23, now acquired by microsoft). Donovan believed in us before yc did. grateful for his support throughout our yc journey. 🐐

p.s. @onethomaschan wanted to "brand awareness-maxx" with @LanternAI__ logo

[Justin Tang @JustinTang29]

yc founders, if u're already in sf, come join us!

https://t.co/K6rMHUUHLW

[Thomas Chan @onethomaschan]

i'm hosting a dim sum w/ @rhobusiness before yc fall batch kicks off 🥟

on 9/6, in SF
details + rsvp: https://t.co/k2d3Z0nvWV

come join us! :)

[Justin Tang @JustinTang29]

@hackernews looking forward!

[Justin Tang @JustinTang29]

looks like we're the first recruiting startup in yc's f26 batch 👀

@onethomaschan
@LanternAI__

[Justin Tang @JustinTang29]

moved to san francisco just in time to make it to @ycombinator's first comedy night! great to meet and mingle with new ppl. thanks for hosting! @legalos @MonacoGTM

[Thomas Chan @onethomaschan]

three months ago, i got rejected from yc (landed on top 10% so we were rlly close)

today, @LanternAI__ is published in the yc directory

gotta celebrate the "small" wins :>

yc f26, we’re just getting started!

[Justin Tang @JustinTang29]

@onethomaschan i only look at github or past projects when hiring. always turn out to be some of the most cracked engineers I worked with.

[Thomas Chan @onethomaschan]

i got into YC on my third try, and it was actually the one i spent the least time on.

@JustinTang29 and i had been shipping almost every month, so every application was different.

the overcorrection is thinking you need a perfect application. make something people want and the it writes itself 🙌

[Alexander Panfilov @kotekjedi_ml]

We can finally talk about it:

We found a way to extract hidden reasoning of frontier models using a vulnerability in the APIs of every frontier AI company.

We verified that our reasoning token count matches billed API thinking tokens 1:1 for most of the prompts we queried.

[Thomas Chan @onethomaschan]

W26 - Rejected
S26 - Rejected
F26 - Accepted

Thanks @ycombinator for taking a bet on @LanternAI__ :)

[Justin Tang @JustinTang29]

@onethomaschan @ycombinator @LanternAI__ statistically speaking, our yc acceptance rate is 33%. above market.