It’s 9:14 on a Tuesday morning and my Slack has four separate messages asking the same question. A founder wants to know if Instinct is worth switching to. A reader forwarded the TechCrunch piece about the $350M round at a $2.5B valuation. Someone else linked the WSJ writeup calling it the latest viral AI assistant rocketing across Silicon Valley. And a former coworker just sent a screenshot with no context and the word “thoughts?”
My honest thought: I have no idea yet, and neither does anyone else sending me these messages.
That’s an uncomfortable thing to say in a week when the numbers are this loud. But this is a toolkit review site, and the whole point of reviewing toolkits is that you have to actually use them. A funding announcement tells you what investors believe. It does not tell you whether the thing works on your codebase, your data, your team’s weird internal conventions.
What the round actually tells us
$350M at $2.5B is real money and a real signal. It means people with access to private metrics looked at the growth curve and decided it justified that price. That’s information. It’s just not the information most people think it is.
Here’s what it reliably tells you:
- Instinct is growing fast enough that sophisticated investors bet nine figures on the trajectory continuing
- The company will have runway to hire, scale infrastructure, and survive a rough patch
- The product almost certainly won’t disappear in six months, which matters more than people admit when you’re picking tools
Here’s what it doesn’t tell you: whether the output quality holds up on your actual work, what the pricing looks like once the growth-at-all-costs phase ends, how it handles edge cases, whether the API is stable, or what happens when you need support at 2am.
The pattern is the story
Look at the surrounding headlines from the same stretch. Higgsfield raised $400M in a Series B that quadrupled its valuation in eight months to $5.4B. Parallel Web Systems hit $2B five months after its last big raise. Ricursive, an AI chip startup, reached $4B two months after launching.
Two months. That’s not a typo, and it’s not an outlier anymore. Valuations in this space are compounding on a timeline that has nothing to do with how long it takes a tool to prove itself in production. A team evaluating software properly needs a quarter at minimum. Some of these companies have gone from launch to multi-billion valuation in less time than one honest pilot program.
This creates a specific problem for anyone trying to make a decision. The funding cycle now moves faster than the evaluation cycle. By the time you’ve finished testing a tool carefully, a competitor has raised twice as much and the conversation has moved on. The pressure to skip the testing and just go with whatever is winning the funding race is enormous.
Resist that pressure
I’ve watched teams adopt tools because of momentum and then spend a year unwinding the decision. The switching costs are never in the marketing material. Data migration, retraining, rewritten integrations, the institutional knowledge that evaporates when you drop a workflow people had finally gotten good at.
Viral adoption is a signal about ease of onboarding, not about depth. Tools that spread fast usually spread because the first five minutes feel great. That’s a genuine achievement and a real product skill. It’s also the part that’s easiest to optimize and least predictive of whether you’ll still be happy in month nine.
What I’d want to know before recommending Instinct to anyone:
- How does it perform on tasks where being 85% right is worse than useless
- What the actual cost curve looks like at team scale, not solo scale
- Whether the model behavior stays consistent across updates or shifts under you
- How much of the experience depends on prompting skill you’d have to teach everyone
- What the exit path looks like if you decide to leave
None of that is answerable from a press release. All of it is answerable in about six weeks of deliberate use.
What I’m doing about it
I’m getting hands on Instinct and running it through the same tests I run on everything else. Same tasks, same comparisons, same attempt to break it. When I have something to report, I’ll report it, including the parts that don’t flatter anyone.
Until then, my recommendation is boring and I’m fine with that. If your current setup is working, keep working. If you’re curious, run a small pilot on non-critical work and take notes. And treat the funding numbers as what they are, which is a story about capital markets that happens to have a software product attached.
The money is real. The verdict isn’t in.
đź•’ Published: