To be blunt, I would rate it a 1, a 0 if I could. I would stop reading after the third sentence. FWIW, humanizers do not work. Can they make AI text sound better - yes, but only mildly. Also, they won't trick good ai detectors, and the resulting text is still cumbersome to read.
This is what I got first time with a humanizer I built, still sounds terrible:
"Yeah, an LLM gets you the 10-minute demo now.
The expensive part is everything after, once real users depend on it. You end up maintaining editable state, templates, email-safe HTML, export pipelines, asset handling, permissions, custom blocks, and live previews. That takes months.
Structure makes AI output usable here, and it gets more important as the workflow gets automated. That's the idea behind Unlayer Elements. The LLM generates real editable components that map back into the actual email or page, and you keep editing what it makes like anything else in the tool.
I'd prototype with the LLM. I'd bring in Unlayer once that content layer becomes real infrastructure you don't want to own yourself."
While that may be true, in this instance the whole thing is 100% AI generated and this is not someone who has just adopted the mannerisms of AI.
For instance, the usage of "Totally fair" and "The hard part", the usage of colons, which humans would use far less frequently, using the rule of threes (editable email, web page or document), etc.
Anyway, it's just lazy. This is someone trying to sell their product. They should be able to speak about it without using the exact response from an AI
London has incredible AI talent but yeah, unfortunately most of them work on American tech. In another life Deepmind would not be owned by Google but alas.
Yeah I feel the same way. Wonder when/if we'll get continual learning from these models. I feel like they are smart enough already but their lack of real memory makes them a pain to deal with.
Their annualized revenue run rate is on track to surpass $6 billion by the end of 2026 so it's not ridiculous for them to be valued at $60 billion at some point. Also worth noting that if they do get access to SpaceX compute, they could start pretraining their own model. Composer is good but its built on top of Kimi 2.5.
Where does it say we recommend you work with scammy low-quality auditors? They say that they use third party audit firms that are used by other compliance companies.
They likely barely had a product when they applied to YC. It's more interesting as to why this wasn't discovered (if it is even true) when they were raising their Series A.
I'm always interested to understand - what constitutes a basic ChatGPT wrapper?
Is Legora, which is doing very well, a basic ChatGPT wrapper? Because if you don't view it as one, it certainly started as one.
Feels pretty easy to mitigate against. If a user deselects "allow email sending", then you can just remove that as a possible tool-call so it becomes impossible.
As someone that uses Twitter quite a lot for consumption, I actually think it's great. I've learned so much on Twitter (and yes, I'm aware there is plenty of disinformation), and it's also extremely entertaining. Maybe I've just used it long enough so I my feed is quite curated.