Vue normale

  • ✇404 Media
  • Substackers Say New AI Detection Tool Is a ‘Witch Hunt’
    Substack contributors are rejecting the company’s decision to flag AI generated writing on the newsletter publishing platform, with several creators saying the new AI detection features are a “witch hunt.” Some Substack users who use AI to help them write or edit articles say the new policy unfairly discriminates against them, while other users who don’t use AI in their writing at all are worried that Substack will mistakenly flag their writing as being AI generated and tarnish their reputation.
     

Substackers Say New AI Detection Tool Is a ‘Witch Hunt’

28 juillet 2026 à 10:53
Substackers Say New AI Detection Tool Is a ‘Witch Hunt’

Substack contributors are rejecting the company’s decision to flag AI generated writing on the newsletter publishing platform, with several creators saying the new AI detection features are a “witch hunt.” Some Substack users who use AI to help them write or edit articles say the new policy unfairly discriminates against them, while other users who don’t use AI in their writing at all are worried that Substack will mistakenly flag their writing as being AI generated and tarnish their reputation. 

“I’m not going to apologize for using AI in the creation process. I wrote for 20 years without AI, I could do it again if I wanted to,” Mack Collier, who has a small Substack called Backstage Pass, wrote. “My output would fall, my posts would be less structured, and it would be more obvious that they needed a good editor. Using AI improves the overall quality of my writing. That’s why I use it.”

“These detectors are notoriously, wildly inaccurate,” Alice Lemee, a ghostwriter and digital writing coach, said in a video she made about Substack’s new AI detection features. “All it takes is one false accusation for a writer to have their reputation almost irreversibly tarnished.”

Substack announced the AI detection features on Tuesday in a post from CEO Christ Best, who said that “It’s getting harder to tell what’s real on the internet” and that “when content made by no one takes over parts of the internet that are supposed to be human, it pollutes the commons and makes it hard to discover and hear human voices.”

Best explained that Substack has partnered with the AI detection tool Pangram, which is now built in to Substack and allows any user to scan a piece of writing. Pangram then produces a percentage-based score determining how much of the writing was AI or human generated. (Disclosure: Pangram previously bought an ad on 404 Media). We’ve covered research from Pangram, or research that relies on its AI detector before, showing how AI generated writing is flooding every corner of the internet. 

As we’ve previously reported, AI detectors are not perfect, and Pangram itself is not immune to labeling human content as being AI, and labeling AI as human. Max Spero, the CEO of Pangram, recently told us that the company is constantly working on minimizing errors, and that it estimates its false positive rate at roughly one in 10,000.  

“We have now built the mirror image. Pangram is a real machine, trained on vast amounts of text to make a probabilistic guess, and we have pointed it at your writing to work out whether a person is hidden inside,” Sam Illingworth, a professor and the author of Slow AI, said on his Substack in a post titled “Substack’s AI Detector and the Return of the Witch Hunt.” “To decide if there is a human on the other end, Substack asks a machine.”

Substack also added a way for users to add a “How I make this” statement, where they can explain their writing process and disclose if they use AI. 

“We’re not against people using AI to assist their work, and we think people should be free to choose which tools they use to express themselves,” Best said. “Some on the platform are human-writing maximalists, others are publicly exploring the frontier of using AI tools while sweating the details to make work they stand behind. We use AI all the time at Substack to write software, do research, and build product features like clipping, translations, and more. But people should know what they’re getting.”

“I think the best way for Substack to address this is to invite further dialogue,” Illingworth told me. “I actually really like the new feature that lets authors tell their readers how they construct their newsletters using AI, and this is the kind of dialogue they should encourage rather than ascribing a score that we know to be incorrect. My worry is that not only do these AI detectors generate false flags for non-native English speakers and neurodiverse writers, but they also remove any opportunity for dialogue because they start from a position of suspicion. Whereas what we need to be doing is developing opportunities for trust between readers and authors.”

"The tools we've introduced today are intended to increase transparency and give readers more context, not to prohibit or penalize AI-assisted writing. They do not impact discovery on the platform,” a Substack spokesperson told me. “We're also encouraging Substack publishers to add a ‘How I make this’ statement, where they can explain their process directly, including how they use (or don't use) AI, and set expectations for readers. Creators can disable detection on their posts pre- and post-publication, as well as report and remove scans on their own work that they believe are mistaken. Find more information on how the features work here.”

Substack’s attempt to detect and label AI generated content was also celebrated by many readers who are exhausted by AI slop and other internet platforms that can’t or don’t want to label it. I just published story about how there is so much unlabeled AI slop on Spotify that people are now creating their own sites to detect and label it without the company’s help.  

But the backlash from some Substack users shows that companies can’t just flip a switch if they want to ban or give users the ability to filter out AI content from their feeds. As the flood of AI writing and images we see online and the real world every day makes clear, generating slop is easy. What we’re going to do about it is still an unsolved problem.

  • ✇404 Media
  • Spotify's AI Problem Is So Bad Random People Are Stepping In to Track the Slop
    Slime Dot, a young R&B artist from Las Vegas, currently has over 100,000 monthly listeners on Spotify. In an interview with GQ in May, after sharing an image of herself posing with Drake on Instagram, Slime Dot directly denied accusations that she was an “AI artist,” saying “The truth doesn’t matter these days. People are always gonna try to explain something they don’t fully understand and believe what they want.” GQ originally credulously published the interview and said it was “debunki
     

Spotify's AI Problem Is So Bad Random People Are Stepping In to Track the Slop

27 juillet 2026 à 09:25
Spotify's AI Problem Is So Bad Random People Are Stepping In to Track the Slop

Slime Dot, a young R&B artist from Las Vegas, currently has over 100,000 monthly listeners on Spotify. In an interview with GQ in May, after sharing an image of herself posing with Drake on Instagram, Slime Dot directly denied accusations that she was an “AI artist,” saying “The truth doesn’t matter these days. People are always gonna try to explain something they don’t fully understand and believe what they want.” GQ originally credulously published the interview and said it was “debunking the AI rumors.” Eventually, GQ added a note to the end of the article conceding Slime Dot was an AI avatar, but said the “talent behind the music remains undeniable.”

This confusion could have been avoided if Spotify had done what many of its users and critics have been begging it to do since AI music generators like Suno and Udio made it trivial to flood the internet with AI music. Spotify could flag Slime Dot and other AI-generated music on its platform as such. Other platforms, like YouTube, already do this. 

Since Spotify doesn’t do that, users can visit SoullessMusic.com, a database for “AI artists hiding on Spotify. No bands, no studios, no soul, just machines and melody.” There, Slime Dot is listed as “almost certainly AI” based on an analysis of three of the tracks and the fact that Slime Dot released such a large number of tracks in a short period of time. Alternatively, users could go to SlopTracker, where they can upload files or copy/paste links to Spotify tracks to run them through a tool which will tell them, with varying degrees of confidence, whether a song was AI-generated or not. SlopTracker found that Slime Dot’s track “Fully” was 95 percent likely to be AI generated by Suno. 

“I didn't know what's AI and what's not,” Graeme Fulton, who created SoullessMusic, told me. “And then I'm on Instagram Reels and I see loads of artists unhappy saying some artist has popped up with their song and just copied it. And there's loads of different cases where they [AI artists] would copy an artist's look and appearance […] sort of just stealing from artists.”

Last year, I wrote a story about an exceptionally bad example of what Fulton is talking about. A scammer used Spotify to publish AI-generated music under the name of a real, dead musician, seemingly in an attempt to syphon money from his popularity via streams. After I published that story I heard from several (living) artists who said the same thing happened to them. A similar case involving a Danish jazz musician was covered by Danish publications in April. 

Another reader recently told me about a popular Spotify artist that was publishing hours of Iranian jazz every week. The channel, Qajar Jazz, which popped up in December of last year, has 29,579 monthly listeners, gained more popularity after the start of the U.S. war with Iran and the channel making vague allusions to resilience and preserving culture in video descriptions and comments on YouTube. Unlike Spotify, Qajar Jazz videos on YouTube are labeled as being AI-generated in the fine print, though it’s unclear if the channel tagged itself as such or if YouTube did. Both SoullessMusic and SlopTracker detect it as being AI generated. 

Qajar Jazz did not respond to a request for comment. 

Many of the YouTube comments on Qajar Jazz videos indicate that people think the music is made by real people. In addition to creating a database of AI generated music on Spotify, SoullessMusic also collects social media posts from people who say they were similarly tricked by AI generated music. Some posts are from musicians who lament the fact that they are now competing with algorithms that can produce thousands of tracks with no effort, and others are from artists whose music was directly ripped off by AI artists

Fulton said he started looking around on GitHub for open source AI music detection tools, and ended up building a new, open source detector that relies on open source models called SONICS, the lofcz vocoder fakeprint detector, and other open data sets. SoullessMusic also  scans a track’s metadata for any tags that might tie it back to a specific AI music generator, and Fulton says he also has a script that scans Wikipedia for existing entries on an artist. Users can submit tracks from Spotify to the site they suspect are AI generated, and Fulton reviews them manually before adding them to the database. Despite all of this, he admits the system is not perfect. 

Both Soulless Music and SlopTracker use AI and other automated tools to detect AI music. As we’ve written previously, using AI detectors to detect AI content is an inherently flawed process that can lead to false positives. When I asked him how he handles cases of false positives, Fulton said “it's not ideal when that happens, but it's going to happen because it's still quite tricky to check what's AI and what’s not.”

Spotify’s lack of transparency and unwillingness to label AI-generated music makes it impossible to track how much money its making on the platform. Last year, the company announced that it would it would AI disclosures on AI generated music, but I haven’t seen it on any of the dozens of tracks I reviewed for this story. Udio and Sudo both include inaudible digital watermarking in their generation, which in theory could make it easy for Spotify to tag these tracks. 

"We're employing a layered approach that combines enforcement, artist controls and greater transparency," a Spotify spokesperson told me after this article was first published. "Over the past year, we've introduced policies targeting harmful AI-related behaviors like spam, impersonation and deceptive content, alongside new tools that give artists more control and listeners more context. These include Verified by Spotify, which helps listeners identify authentic artist profiles; AI Credits, where artists disclose when and how AI was used in creating their music, with tens of thousands of AI credit disclosures now being submitted each day; and Artist Profile Protection, which gives artists more control over what appears on their profiles. We are continuing to build on these efforts.”

Spotify also pointed me to several tracks it had flagged as AI on the mobile app, like Better Times by Mike Mana.

Soulless music attempts to quantify that number based on publicly available streaming data. The small number of AI artists tracked on SoullessMusic generate an estimated $5.7 million a year, with the most popular AI artist, mikeeysmind, generating $1.5 million annually alone. In April, Deezer, a Spotify competitor that attempts to tag all AI-generated music on its platform, said 44 percent of all new music uploaded to its platform is now AI generated. SlopTracker, which attempts to track AI generated music on Spotify that is featured on official Spotify-curated playlists, that those artists are “draining” $0.1188 per second from real artists who could be making the music instead.

  • ✇404 Media
  • A Court Reporter Submitted AI-Generated Errors in Official Court Transcript, Judge Says
    A judge caught a court reporter making AI-generated errors in a court transcript, and put court reporters everywhere on notice for their use of AI. In a memorandum decision concerning a case about a man who sold drugs to another man who overdosed and died, filed on July 23, Judge Paul Felix wrote in a footnote of the decision that a transcript contained errors that looked a lot like generative AI. The footnote was spotted by attorney Rob Freund on X.
     

A Court Reporter Submitted AI-Generated Errors in Official Court Transcript, Judge Says

23 juillet 2026 à 17:41
A Court Reporter Submitted AI-Generated Errors in Official Court Transcript, Judge Says

A judge caught a court reporter making AI-generated errors in a court transcript, and put court reporters everywhere on notice for their use of AI. 

In a memorandum decision concerning a case about a man who sold drugs to another man who overdosed and died, filed on July 23, Judge Paul Felix wrote in a footnote of the decision that a transcript contained errors that looked a lot like generative AI. The footnote was spotted by attorney Rob Freund on X.

  • ✇404 Media
  • Patreon Lays Off 20 Percent of Its Workforce
    Patreon laid off 93 employees, totaling 20 percent of its workforce on Thursday morning, according to an email to creators and staff from CEO Jack Conte.In a message sent to everyone on the platform signed up as a creator, and posted to the site, with the subject line “A Painful Update about our Team,” Conte wrote that the business is “healthy and strong” and that the core business of Patreon is not changing. Conte included the email sent to Patreon employees in the message. In the email, he
     

Patreon Lays Off 20 Percent of Its Workforce

23 juillet 2026 à 11:24
Patreon Lays Off 20 Percent of Its Workforce

Patreon laid off 93 employees, totaling 20 percent of its workforce on Thursday morning, according to an email to creators and staff from CEO Jack Conte.

In a message sent to everyone on the platform signed up as a creator, and posted to the site, with the subject line “A Painful Update about our Team,” Conte wrote that the business is “healthy and strong” and that the core business of Patreon is not changing. 

Conte included the email sent to Patreon employees in the message. In the email, he wrote that the company is undergoing both a “workforce reduction” and is “changing our organizational structure and how we work.” Specifically, he wrote, this means “we’re flattening the organization, refocusing teams on our top priorities, and evolving key aspects of our operations to make us faster at adapting to change.”

Conte is careful in this email to both express that he doesn’t view AI as a replacement for the human creativity the platform is built on and profits from — devoting a section of the email to saying as much — and also that AI is fundamentally “transforming” the tech industry, noting that it has an impact on how the company operates. 

“To be clear about the impact of AI on today’s decision: we are not making the above changes because we believe AI replaces humans,” Conte wrote. “The more we have learned to use these new tools, the clearer it has become that they are not substitutes for the creativity, judgment, detail orientation, or craftsmanship that our teammates have in spades, nor do they replace the desire for human connection that all of us cherish so deeply. That’s my personal opinion, but more importantly, it’s the foundation of Patreon’s strategy: our product vision and business are both predicated on the value of human creativity and human connection. AI has fundamentally transformed the tech industry, though, including how we work, how we build products, how we communicate, and more. That does have an impact on how we operate and organize.”

Earlier this month, Patreon announced that it’s partnering with Cloudflare to block crawlers from stealing creators’ work to train AI models. “I HAVE A KICKASS PRODUCT UPDATE FOR YOU ALL!” Conte wrote in a post on Instagram. “This is live and happening at the network level on all posts published on Patreon.” The company later elaborated on the partnership in a blog post.

Included in the internal email shared publicly are severance details for laid-off workers, including 16 weeks of pay, remaining on payroll through the company’s August 20 vesting date, one additional week of pay for every full year worked at the company, additional cash payments for recent hires who haven’t reached their one-year cliff, healthcare coverage through the end of the year for eligible employees and families, and a $1,500 stipend to replace their company laptops.

  • ✇Climb to the Stars
  • Thoughts on Letting AI Write For You [en]
    [en] I was planning to write another blog post. I may still write it – or not. But as I opened my computer, I had a pile of open tabs. I’d clicked on a bunch of articles of Daniel Miessler’s blog for possible further reading. Daniel is the guy behind LifeOS (previously PAI), a personal AI framework (infrastructure) I have written about and spent many (many!) hours battling with since April. It had been a pleasant surprise to discover Daniel’s blog was active, and even more that it was readable,
     

Thoughts on Letting AI Write For You [en]

22 juillet 2026 à 14:53
[en]

I was planning to write another blog post. I may still write it – or not. But as I opened my computer, I had a pile of open tabs. I’d clicked on a bunch of articles of Daniel Miessler’s blog for possible further reading. Daniel is the guy behind LifeOS (previously PAI), a personal AI framework (infrastructure) I have written about and spent many (many!) hours battling with since April.

It had been a pleasant surprise to discover Daniel’s blog was active, and even more that it was readable, as most of what I’ve seen “from him” these last months is in the GitHub repo or PAI/LifeOS documentation and rather AI-sloppy, sometimes even to the extent that it sounds like his DA (Digital Assistant) posting on his behalf. (I say this with all due respect, the thinking behind PAI/LifeOS is genius.)

Not so with the blog, clearly well-written by a human (see, by the way, his AI Influence Level system, something in the same vein as AI Labels).

So I fell in the trap of looking at the topmost open tab before closing it, and I started reading, and I was even more pleasantly (ephorically?) surprised to read this:

I value text because I see it as one step away from thought.

I believe thinking is the one thing we should be careful not to outsource, and I worry what this idea smuggles in is a major step toward making our creations opaque to humans. Not just AI’s creations, but ours as well.

The reason I value Paul Graham so much is because of the idea compression work that goes into writing super clean prose. It’s difficult to write clearly because it requires thinking clearly.

Daniel Miessler, “Text is Thought, and Thought is Holy

I worry that if we vibe-think to AI and have it spit out amazing HTML, we’re instantly disconnected from the idea. Like where did the idea go? It started as vibes and got put through a woodchipper and turned into someone else’s HTML.

I see this happen so much. A lot of writing is a way of thinking, it’s not “path to an output”. And I think that when thinking is missing in the process, we feel it, which is why AI Slop and even many “decently AI-generated texts” are so distasteful to read.

Because the “thought” they contain is not coherent, not organic, and we cannot access it comfortably through the text we’re reading. We need to do extra work to try to get at it – and sometimes it’s not even there. Just like “over-optimised” administrative processes end up outsourcing a large part of the work they’re supposed to do to the people they should be serving, “over-optimised” text production outsources part of its work to the reader.

We could maybe apply this analysis to badly constructed, hastily-written reports whose sections have been copy-pasted to death and which lack overall integrity and structure.

Jens-Christian wrote about this too in “Keep Your Voice“.

Daniel again:

I just feel like if you didn’t put the hard thinking and writing work into the original idea, and then maintain it in a format that’s easy for humans to read and edit, then you have somehow surrendered something Holy to the machines. I say this as a total AI maximalist.

This reminds me of two things that came upon my radar recently.

The first is this post by Alex Hillman about how generative AI has broken the social contract by upending the balance of effort from producing to consuming content. What used to be expensive (writing, creating a song, shooting a video, etc.) is now super cheap, in terms of effort (if not tokens). Sure, we used to have content farms and the like, but it was still a lot of manual labour. Now I can feed a one-line prompt to an AI model, leave the computer running and come back to a whole book of content. Sloppy, but a lot of words.

Over time, the world we live in has shifted more and more in the direction of “too much content, not enough brain time to absorb it”. It’s something humans have been complaining of since before the internet. But like everything, it’s speeding up like a hockey stick. The democratisation of online publishing, and social media on its heels, has brought in another height of cognitive overload and collective anguish about there being “too much stuff out there”. And now that more and more of this stuff is produced by machines that never need to sleep or eat (except a prompt or two), the overall quality of what is out there is going down as fast as its quantity is going up.

This reminds me somewhat of my thoughts about audio and video versus text, back in the day. If you record a 30-minute video and post it online, it took you 30 minutes, but it will also take each person who watches it 30 minutes. If you spend 30 minutes writing up your message, it will take the people reading it much less time.

Because what has not changed is the finitude of human existence. Not only is our time limited, but also our energy and ability to read, watch, learn, concentrate. This is, now more than ever, our bottleneck. Producing something today is pretty much worthless. Producing more is senseless. What has value is: does it enter somebody’s mind in a meaningful way?

There is a link between this and the AI brain fry explanation that I referenced some time back. AI has no limits on the concurrent projects it can work on, and no limits on how much content it can produce. But we have limits on how much delegation we can supervise, how many open loops we can stand, and how much content we can absorb.

And this brings me to the second thing: understanding cannot be outsourced. Andrej Karpathy now famously said you can outsource thinking, but not understanding. A lot of what AI produces does not help us understand anything. It gives the illusion of opening the door to that, but it’s more often than not just word froth. Reading is about understanding. It’s about meaning. There is no shortcut to that. Only you can learn and understand so that you later know.

When Karpathy says you can outsource thinking: well, some of it. But what we writers-as-thinkers know very well is that the thinking that we do when we write is the kind of thinking that leads to understanding. Saying that thinking can be outsourced is a slippery slope because it drags down a chunk of understanding with it.

Writing is thinking is understanding.

Well, that was the first open tap. 27 more to go. (Just kidding. I’m going to close them. I might write my Paléo Festival blog post tomorrow.)

  • ✇Climb to the Stars
  • Less Input [en]
    [en] In the spirit of shortening things, I’m taking a few moments during my lunch break to share some thoughts I’ve been having recently. Various things have contributed to these thoughts: my permanent struggle with “too many ideas” and “too many things I want to do” (which predates my accident, but is now exacerbated given my reduced energy) pondering on how to manage my energy (already underway since my accident, but now fed by the occupational therapy programme for energy management t
     

Less Input [en]

8 juillet 2026 à 07:37
[en]

In the spirit of shortening things, I’m taking a few moments during my lunch break to share some thoughts I’ve been having recently. Various things have contributed to these thoughts:

  • my permanent struggle with “too many ideas” and “too many things I want to do” (which predates my accident, but is now exacerbated given my reduced energy)
  • pondering on how to manage my energy (already underway since my accident, but now fed by the occupational therapy programme for energy management that I’m in the middle of)
  • my training at IGB, and the two-day course I just did in Paris (the recurring focus is “how something that works for the person at some point becomes the thing that feeds the problem”)
  • my exploration of AI and in particular the framework/project/took that was previously called PAI (now LifeOS) (output is cheap now: feed a genAI model a few lines and it can spit out thousands of words for you).

I’m not going to be able to reconstruct how my ideas around this have shifted in a chronological or well-organised way, but here’s more or less where I’m at: I’m somebody who believes that “knowledge is power”, that “more information is better”, that by learning and analysing and understanding, I can find answers.

That works a lot of the time, for me. It has worked very well for me. But I’m starting to see how this is part of what is trapping me, right now.

At some point in my struggles with my AI infrastructure (trying to get the PAI Digital Assistant up and running correctly, fixing bugs, making sure it learns correctly, setting up workflows for the things I want it to do for me) I realised that it had a built-in bias (the model, most probably) towards “producing more is better”. LLMs are verbose, we all know that by now. The more you feed them, the more words they spit out – but not necessarily the more useful information.

I kept giving instructions for concision. I would provide examples of how to write things up. I would set up guardrails, have it self-correct, hunt for fluff and filler content. At some point I realised that “the system” was just growing and growing in terms of content (the number of words and files that contained the instructions), and the output was not improving – more like the contrary. This is nothing new, right. We know that complex systems balloon up and lose efficiency. And I’ve seen more than once in my dealings with AI that I pretty much always end up spending more time “fixing the system” than actually using it. This, actually, is something I’d noticed about myself in general; but it’s easier for me to rein in when I don’t have an LLM at the other end of the keyboard. So here, it became even more visible.

So here it is. The mantra I keep repeating to people in all sorts of context: “less is more”; “better is the enemy of good” (that’s what we say in French). If I feed my AI system less input, I get less bloated output in the system. I read somewhere (can’t remember the source, probably have it stashed away somewhere) that in the age of generative AI, the bottleneck was shifting from content production to content consumption. What we can “ingest” is the limiting factor, now that we have machines that can spew out words and sentences and paragraphs and essays and reports like there’s no tomorrow. But it’s worthless unless we can read it and understand it and do something with it. Just feeding that AI output to another AI is just going to magnify errors and biases and produce more slop, unless there is a humain mind in charge that understands what it’s doing and what it’s asking.

Early on, in 2024, I remember reflecting at some point that the AI I was using seemed “more ADHD than me”. And what my more recent experiences have helped me understand about myself is this:

  1. My problem, my “slop/bloat” is ideas and things I want to do. I am somebody for whom ideas are cheap. I have ideas all the time. I can come up with stuff I’d like to do all day. My problem (bottleneck) lies in selecting what I run with, and that is a difficult exercise – even more difficult since my accident.
  2. The more “input” I get, the more – just like the LLM – I produce ideas and desires. I read an article, there are 5 more I want to read. I have a discussion on a topic, I want to read more stuff or write articles or create a community around it. I learn about something new, oooh it’s nice and shiny and I want to do it.

Having a large capacity for input and lots of ideas, followed by enough energy to take a handful of them and run with them (OK, frustrating to have to leave so many aside, but at least I’m busy doing something useful/interesting with some of them) has, as I said above, served me very well in life. But trees don’t grow to the sky. At some point, what has worked well becomes the source of the problem.

And I’m realising that the way out of this, at least now, is not better prioritising. First of all, it’s reducing input.

Less input, less ideas, less “oh I want to do this thing”, less slop to sort through, less frustration with everything I’m not doing.

Exactly how to achieve that is still a thought in progress. But that’s what I’ve been thinking of this last week or so.

  • ✇Climb to the Stars
  • La clé du véritable assistant IA: l’infrastructure IA personnelle (PAI) [en]
    [en] Même si je ne suis pas en train de vibe-coder à tour de bras, j’ai bien sauté dans le train côté IA. Depuis deux ans environ, je dis que ce que je ressens concernant l’importance et le potentiel de l’IA générative est quelque chose que je n’ai pas senti depuis ma découverte d’internet il y a 20-25 ans. Et là mes amis, j’ai l’impression de m’être pris un train en pleine figure. Enfin, je le vois arriver depuis un moment, le train, mais là il est sur moi. Mon vieux pote Jens-Christian vie
     

La clé du véritable assistant IA: l’infrastructure IA personnelle (PAI) [en]

17 avril 2026 à 16:01
[en]

Même si je ne suis pas en train de vibe-coder à tour de bras, j’ai bien sauté dans le train côté IA. Depuis deux ans environ, je dis que ce que je ressens concernant l’importance et le potentiel de l’IA générative est quelque chose que je n’ai pas senti depuis ma découverte d’internet il y a 20-25 ans.

Et là mes amis, j’ai l’impression de m’être pris un train en pleine figure. Enfin, je le vois arriver depuis un moment, le train, mais là il est sur moi. Mon vieux pote Jens-Christian vient de me montrer, en live (bon, par visio) à quoi ressemblait son assistant IA et l’infrastructure (PAI) qui le rendait possible. Je vais vous expliquer tout ça en français, ne vous en faites pas, mais sachez déjà qu’en gros, le fameux “assistant IA” qui est véritablement capable de nous aider à organiser nos vacances, choisir un menu pour ce soir, nous briefer pour préparer notre journée, garder le fil de nos multiples projets et évidemment, automatiser notre administratif récurrent, sans juste faire semblant qu’il en est capable – eh bien on y est.

Je précise d’emblée, parce que le dégueulis d’IA – je sais pas comment traduire mieux “AI slop” – envahit non seulement le web, les réseaux, nos conversations WhatsApp et les blogs, que j’écris ce texte avec mes dix petits doigts sur mon clavier. Le jour où je vous refile un truc “écrit avec IA”, je vous le dirai, et je vous dirai aussi “combien écrit avec l’IA” c’est.

OK. Maintenant j’explique pourquoi je suis en train de vous dire que l’ordinateur de Star Trek, c’est déjà aujourd’hui. Enfin avant, un petite digression préalable.

On le sait bien: ce qu’on appelle aujourd’hui “une IA”, c’est en fait un “LLM” (Large Language Model). C’est une intelligence artificielle (un type de programme) qui produit des mots. En très simplifié, c’est du texte prédictif dopé. Le LLM, il ne fait qu’une chose, à la base: il regarde le “contexte” (le texte déjà écrit) et fait une prédiction statistique sur le prochain mot. Et le prochain. Et le prochain. L’IA (le LLM) ne “sait” rien. C’est juste une machine à aligner des mots. Mais ce qui est dingue, c’est que cette machine est capable de produire du texte que l’on reconnaît comme “discours”, et qui nous donne le sentiment d’être en train de parler à une vraie personne. C’est le côté “chatbot” ou “interface conversationnelle”. Ça veut dire que pour faire faire des choses à une machine, aujourd’hui, on peut simplement lui expliquer avec nos mots – pas besoin de cliquer sur tel bouton, donner telle commande, utiliser un langage de programmation. On dit quelque chose, et quelque chose se passe.

Si vous avez déjà l’habitude de fréquenter Claude ou ChatGPT, vous savez comme ça va. On chatte avec, et on est tour à tour bluffé (quand ça marche) et désespéré (quand ça marche pas). On se casse vite le nez sur les limites du truc, il nous dit un truc qui est faux, et on se dit qu’il est bien con et que l’IA, ça ne marche pas si bien que ça. On pense aux rêves qu’on avait de pouvoir dire à notre IA “fais ma déclaration d’impôts” ou “planifie mon projet” (pourquoi pas “fais mon job”, pendant qu’on y est), ou plus modestement, “envoie les factures vétérinaires à l’assurance pour qu’ils remboursent”. L’IA nous promet monts et merveilles mais ne livre pas toujours – d’autant plus si on se contente de lui parler dans une conversation qui s’allonge à l’infini, qu’on ne comprend comment travailler avec la fenêtre de contexte, que les skills ça nous dépasse, qu’on ne maîtrise pas les subtilités du prompting et qu’on n’a pas encore fait le pas d’organiser nos interactions en projets ou d’essayer Cowork. On entend bien ceux qui développent des trucs incroyables à l’aide de l’IA, mais bon, on n’est pas tous développeurs, et franchement les résultats médiocres qu’on obtient nous laissent penser que c’est beaucoup de hype, toute cette histoire. On est d’accord?

Et là au milieu, je suis en train de vous dire (je vous promets j’ai pas fumé) qu’en fait oui, l’IA est bien capable de nous offrir cet “assistant digital” dont on rêve. Je crois franchement qu’on y est. Mais c’est pas “en chattant avec Claude”. Et il n’y a pas besoin non plus d’apprendre à programmer pour y arriver.

La clé, c’est cette infrastructure mentionnée en début d’article: PAI. C’est pas une “alternative” à Claude ou ChatGPT ou Gemini. C’est un système qui se construit dessus. Une collection d’instructions pour IA et de scripts (comme des mini-programmes) qui vont tourner sur l’ordinateur.

Comme dit plus haut, une IA ne “sait” rien. On peut lui donner une liste de tâches en début de conversation, en rajouter 3 nouvelles, parler de la pluie et du beau temps, et lui redemander les tâches 10 minutes plus tard, et il y a fort à parier qu’il y aura une erreur dans la liste. Qu’elle reconnaîtra de bonne grâce quand on la lui fera remarquer. Par contre, si au début de la conversation on lui demande de créer un fichier contenant cette liste, et que deux jours plus tard on lui demande nous lire la liste qui est sur le fichier, là, ça marche. Suivant quelle est votre expérience dans l’utilisation de l’IA, vous avez peut-être fait ce constat de vous-même: c’est vachement plus fiable de faire mettre des infos dans des fichiers, de faire écrire des scripts pour extraire de l’information d’un document ou d’un tableau de données. Et d’ailleurs, aujourd’hui, votre “Claude chatbot” le fera spontanément ou vous le proposera, suivant ce que vous lui demandez.

Donc: l’assistant IA de nos rêves, c’est pas “juste” une IA, c’est surtout des tas de scripts et des fichiers contenant des informations. Et bien sûr quand même une IA, pour interagir avec nous (la fameuse “interface conversationnelle”) et faire évoluer le système en fonction de nos besoins.

Vous êtes encore là? Soyons un peu concrets. PAI c’est donc un truc qu’on télécharge sur son ordi et qui une fois lancé, va premièrement installer ce dont il a besoin (y compris Claude Code si on ne l’a pas déjà). Ensuite, quand on le lance, il va nous prendre par la main (via l’interface conversationnelle, donc en discutant déjà avec l’IA) pour le configurer et démarrer avec. Oui, il faut taper un truc dans la ligne de commande, il y a un premier pas ou deux un peu geek, mais après, c’est “juste du blabla” comme on a l’habitude de faire normalement avec une IA.

Ce qui fait que PAI c’est pas “juste une IA”, c’est que c’est un système prévu pour créer ce dont vous avez besoin de votre assistant. On ne vous livre pas une voiture: on met à votre disposition tout ce qu’il faut pour concevoir et produire la voiture qui correspond exactement à vos besoins – les ingénieurs, les designers, les mécanos, l’usine de production, et aussi le chef de projet qui va vous prendre par la main pour vous aider à décrire la voiture dont vous avez besoin. PAI c’est ça.

L’infrastructure comprend un système et une structure pour stocker des infos de contexte vous concernant – pas juste la façon dont vous aimez que l’IA vous parle, mais aussi vos valeurs, vos domaines d’expertise, vos préoccupations et buts dans la vie. Et pas sous forme de laborieux champs texte à remplir. L’IA va vous interviewer, ou dans mon cas, avaler mon blog, et en tirer les informations pertinentes. Ou pas, si on veut pas. L’infrastructure comprend aussi des processus: en particulier, comment développer un système dont vous avez besoin.

Un exemple en guise de démo (on a fait ça en direct avec Jens-Christian et son assistant pendant notre visio): un truc dont je rêve, perso, ce serait un système qui est au courant de ce que j’ai dans mon frigo et mes armoires, de ce que j’aime manger et cuisiner, et qui puisse me dire “ah ben ce soir, tu pourrais te faire ceci ou cela”. Pour de vrai, c’est une grosse charge mentale chez moi ce genre de truc. Donc hop, on soumet cette idée à l’assistant. J’ai deux ou trois idées en plus: si je donne à l’assistant mes tickets de caisse de courses, il pourrait savoir ce que j’ai acheté comme nourriture. Et si je lui montre ce que je mange ou me cuisine, il pourrait en déduire ce que j’ai utilisé. Peut-être de temps en temps il faut lui faire une photo du frigo ou de l’armoire pour vérifier si son inventaire est à jour. Et il pourrait aussi avoir à dispo une collection de recettes que j’ai faites, et d’idées-repas (j’ai commencé à compiler ça, manuellement). On donne donc ces infos à l’assistant, et on observe. Honnêtement, c’est là que je me suis retrouvée sur le cul.

PAI comprend donc un processus, ou une méthode, pour approcher ce genre de demande: clarifier à quoi ressemblerait le système fini, puis, selon une logique similaire au “rétro-planning” (du reverse-engineering, en fait) il produit un concept, les différentes parties du système à créer, évalue de quelle façon les créer (scripts, IA, quels outils sont à disposition, parce que bien sûr il sait quels outils on a sur l’ordi et ce qu’on a l’habitude d’utiliser), à quoi ressemblerait une version minimale fonctionnelle du système (le “MVP”=minimum viable product du jargon dev/business), et (après feu vert bien entendu) produit un prototype. En dix minutes, on avait un prototype interactif à qui je pouvais demander quoi manger ce soir, après avoir rempli une liste bidon de ce qui était dans mon frigo – et surtout, un plan pour développer le reste du système.

Avant aujourd’hui, j’avais prévu d’écrire un article sur ce que j’étais en train de faire avec l’IA en ce moment. Je me suis mise à Cowork, et je suis en train de créer un système qui m’aide à garder le fil de mes multiples projets en cours. Pas juste à savoir où j’en étais et quelle était la prochaine chose à faire quand je reprends un dossier trois semaines après l’avoir laissé en plan, mais aussi à avoir une vue d’ensemble du “portfolio de projets”. Dans mon système, il y a un fichier dans lequel je note les unes derrières les autres toutes les idées qui me passent par la tête, et ensuite il y a un script (et un peu d’IA) qui avale ce fichier et qui dispatche chaque idée ou info dans une sorte de boîte de réception pour le projet en question.

Exemple (parce que c’est mieux les exemples). J’ai trois tonnes de fichiers d’archives pas mal en bordel sur deux disques durs externes. Ça prend beaucoup de place, et je sais qu’il y a du contenu dupliqué, mais pas où et lequel. Je suis donc en train d’utiliser Claude pour m’aider à faire ça de façon méthodique. Donc ça c’est un de mes projets avec Claude. Mettons que pendant qu’on est en train d’avancer dans ce projet, je me retrouve dans le dossier qui contient les sauvegardes des vidéos live que j’ai faites sur Facebook. Il y a celles des chats diabétiques, mais aussi des vidéos sur d’autres sujets et dont je voudrais faire un article dans le blog. Comme je suis en grande conversation avec Claude, je mentionne ça, et peut-être que j’évoque les 3-4 articles que j’ai en tête, pour qu’il en prenne note. Il va noter ça (parce que je lui ai donné des instructions dans ce sens à l’aide d’un skill) dans le fichier “de sortie” du projet, prêt à être importé dans le fichier d’entrée du projet “idées pour le blog”. Ou alors, je suis en train de faire à manger et une idée de génie pour l’aménagement de mon balcon me traverse le cerveau: je la note rapidement dans mon fichier central, sachant que Claude rangera cette info au bon endroit.

Donc créer ce genre de système, avec Cowork, c’est possible, mais je fais beaucoup manuellement. Je micro-manage beaucoup. Je dois être directive, parce que sinon Claude me dit “ouais ouais c’est bon je m’en charge, c’est super simple” et en fait… non.

Avec PAI, il y a déjà des instructions et des “compétences” (skills) dans l’infrastructure exprès pour que créer ce genre de système se fasse bien et simplement. C’est donc une boîte à outils, accessible via le chat, qui nous permet de “forger” le système dont nous avons besoin – créer l’assistant qui va simplifier les tâches à peu de valeur ajoutée de notre quotidien, ou celles qui sont difficiles pour nous, qui va nous permettre d’avoir plus d’énergie à disposition pour les choses qui comptent pour nous.

Je vais gentiment prendre congé de vous pour aujourd’hui, parce que ça devient un article-fleuve (qui plus est pas super bien structuré et clair peut-être) et que mon cerveau n’a plus beaucoup de batterie. On va en reparler quand je sauterai le pas (dès que j’ai un peu de temps à dispo) pour installer PAI chez moi.

Juste deux mots de prudence avant de clore:

  • Claude Code, Cowork, PAI: des outils puissants mais qui présentent également des risques côté sécurité. Ce n’est pas une bonne idée, aujourd’hui, de laisser votre IA contrôler votre navigateur web, de lui donner libre accès à vos fichiers ou à votre e-mail, surtout si vous ne comprenez pas bien les enjeux sécuritaires
  • faire tourner quelque chose comme PAI “coûte” en termes d’utilisation d’IA – tout comme ça coûte de faire produire des images par l’IA, des pages de code, de manipuler des fichiers sur un disque dur, de lire des PDFs de 500 pages… sur un plan gratuit, oubliez; même un plan à 20.-, c’est probablement bien chaud

Quelques liens (en vrac parce que je suis trop raide):

  • ✇Climb to the Stars
  • The Wild Magical World of AI (LLMs) [en]
    [en] End 2024 is when I really started to get going with “AI” (or LLMs – I know saying “AI” is kinda wrong, but there seems to be no good way to escape it right now). I’d been dabbling a bit before that: as a search engine, to help me with my excel formulas at work, or translating an e-mail here or there. But over that Christmas break, I realised it went further than that: helping me manage my tasks; keep lists updated; make sense of my investment portfolio; troubleshooting technical stuff that
     

The Wild Magical World of AI (LLMs) [en]

5 avril 2026 à 10:59
[en]

End 2024 is when I really started to get going with “AI” (or LLMs – I know saying “AI” is kinda wrong, but there seems to be no good way to escape it right now). I’d been dabbling a bit before that: as a search engine, to help me with my excel formulas at work, or translating an e-mail here or there. But over that Christmas break, I realised it went further than that: helping me manage my tasks; keep lists updated; make sense of my investment portfolio; troubleshooting technical stuff that’s above my pay-grade; and build actual systems to do stuff. I didn’t really know what I was doing, but I saw enough to find it really exciting. I was quite busy with work and life during that period, and not finding time to blog much.

And then I had a skiing accident. Everything stopped, but in time, as I started becoming more functional, I turned back to playing with AI some more, doing a little more vibe-coding too. Excitement grew. But I was also hitting limitations, fed up with the sycophancy, context bloat, hallucinations and endless rabbit-holes. Oh, and really sick of AI slop. (Really: nobody wants to read your AI slop, people.) But the exciting was still there, I was just biding my time. Integrations, agents and AI-powered browsers showed up, with the security risks that are bundled in. I was tempted but stopped myself. And a few loved ones.

Three articles amongst many that tell cautionary tales. I have more to say, particularly about the brain fry one, but not today.

https://brave.com/blog/comet-prompt-injection/ https://hbr.org/2026/03/when-using-ai-leads-to-brain-fry https://www.theguardian.com/lifeandstyle/2026/mar/26/ai-chatbot-users-lives-wrecked-by-delusion

Recently, like many others, I jumped ship from ChatGPT to Claude. I’d been thinking about it for a while, because I have been working hard on migrating my diabetic cat community from Facebook to Discourse, and had come to the conclusion I would have to code a plugin or component or two, and it seemed obvious that Claude was the better AI assistant for that. Around that time, I read this article:

https://derekhanson.blog/tufte-blocks/

In it, the author describes one key aspect of working with AI that I had understood to be important, but that I didn’t know how to put into practice: using different conversations for different “roles” or aspects of the project. And here I had a real example.

I put that in practice (a bit as an exercise) in applying these instructions to migrate my personal context from ChatGPT to Claude. It was extremely satisfying and a great learning experience. I was itching to get going with Claude Cowork and Claude Code, but still a bit anxious. As I see it, Cowork is like being handed a powerboat when all you’re used to are the free permitless motorboats you can rent by the hour on a sunny Sunday afternoon like this one. You can really get in trouble if you’re not careful. And probably, even if you are.

I asked around a bit and Claire pointed me to this wonderful guide to Cowork. What she also explained to me is that Cowork is sandboxed, so it only has access to the folders you give it access to. That’s reassuring. And just a few days ago, Matt shipped Taxonomist, an AI-app (? what do we call these things?) that he used to cleanly recategorise all his blog posts. Unsurprisingly, categories here on CTTS have been a mess since time immemorial, and one of the things that has been clear for some time for me is that AI can help me clean things up a bit around here – in general, not just the categories. But that felt like a great way to get started seeing what the powerboat can do, with a trusted friend on board to keep me from crashing on the rocks.

So earlier today I installed Claude Desktop, downloaded the guide, and started it in Cowork. And wow. Honestly. It’s wild. My 3TB of archived files going back nearly three decades are now hopeful they will one day be deduplicated and cleaned up. Of course, I hit my usage limit (which is why I’m writing this instead of playing with my new AI friend), so I went to have a look at what something like Taxonomist really looks like under the hood. You know, open the files and read them. And it’s starting to come together in my mind. The powerboat is starting to feel like something I will be able to manage in time.

Of course, I’m not going to point Cowork at my external hard drive right now. There are still a lot of steps until I feel comfortable enough with the powerboat to attempt that. But doors are opening. My diabetic cat community migration feels more manageable. I’m hopeful I can tidy my files and clean up my blog. Come up with a system to make sharing links to the open web as easy as on Facebook. Reboot the blogosphere. And I’m sure other ideas will come along the way. I am also trying to be very, very careful about AI brain fry, as I already have my own concussion brain fry to deal with.

If you haven’t yet started learning how to use AI beyond as a proxy for Google or Wikipedia, really, it’s time to get cracking. I’m going back to Lesson 5 (honestly, just read through that page for starters if you don’t know what to do first), making a note to check out Cursor (I’m using Visual Studio Code for now) and read this article on Teaching AI to Design.

First, however: a nap.

  • ✇The Kyiv Independent
  • Ukrainian drone strike hits major oil refinery in Russia's Krasnodar Krai, HUR source claims
    Long-range Ukrainian drones struck the Ilsky oil refinery in Russia's Krasnodar Krai on July 7, hitting one of the facility's technological workshops, a source in Ukraine's military intelligence (HUR) told the Kyiv Independent. Located roughly 500 kilometers (311 miles) from Ukrainian-controlled territory, the refinery is among the largest in southern Russia, producing over 6 million tons of fuel annually. It is involved in the reception, storage, and processing of hydrocarbons and distributes r
     

Ukrainian drone strike hits major oil refinery in Russia's Krasnodar Krai, HUR source claims

7 juillet 2025 à 08:01
Ukrainian drone strike hits major oil refinery in Russia's Krasnodar Krai, HUR source claims

Long-range Ukrainian drones struck the Ilsky oil refinery in Russia's Krasnodar Krai on July 7, hitting one of the facility's technological workshops, a source in Ukraine's military intelligence (HUR) told the Kyiv Independent.

Located roughly 500 kilometers (311 miles) from Ukrainian-controlled territory, the refinery is among the largest in southern Russia, producing over 6 million tons of fuel annually.

It is involved in the reception, storage, and processing of hydrocarbons and distributes refined products via road and rail. The refinery is part of Russia's military-industrial complex and plays a direct role in supporting Moscow's war effort, the source said.

The Russian regional operational headquarters claimed that "drone debris" fell on the oil refinery.

The strike marks a renewed wave of Ukrainian attacks on Russian oil infrastructure, following a months-long pause since March. On July 1, Ukrainian drones struck the Saratovorgsintez oil refinery in Russia's Saratov Oblast.

Kyiv has targeted dozens of refineries, oil depots, and military-industrial sites since the start of Russia's full-scale invasion in 2022. Winter drone attacks forced at least four Russian refineries to temporarily shut down.

This is the second known strike on the Ilsky refinery. Ukrainian drones, operated by the Security Service (SBU) and Special Operations Forces (SSO), previously targeted the facility on Feb. 17, causing a fire.

Krasnodar Krai, a strategic region along Russia's Black Sea coast, has increasingly come under Ukrainian drone attacks as Kyiv extends the range of its strikes deep into Russian territory.

HUR publishes Russian military order, claims proof of Moscow increasing military footprint in Armenia
HUR first made the claim on July 5, saying Russia was increasing its forces at the Gyumri base to exert greater influence in the South Caucasus and “destabilize the global security situation.”
Ukrainian drone strike hits major oil refinery in Russia's Krasnodar Krai, HUR source claimsThe Kyiv IndependentChris York
Ukrainian drone strike hits major oil refinery in Russia's Krasnodar Krai, HUR source claims
  • ✇The Kyiv Independent
  • Drones reportedly attack Russia's Black Sea fleet
    Drones attacked Russia's Black Sea Fleet at the port of Novorossiysk in Krasnodar Krai overnight on July 6, the Russian media outlet Astra reported.Ukraine has not officially commented on the reported strikes, and the Kyiv Independent could not independently verify the claims.An air alert was sounded in the city for several hours, and air defense was active. The consequences of the attack are still being determined, according to Astra.The media outlet also published footage purportedly showing a
     

Drones reportedly attack Russia's Black Sea fleet

6 juillet 2025 à 10:14
Drones reportedly attack Russia's Black Sea fleet

Drones attacked Russia's Black Sea Fleet at the port of Novorossiysk in Krasnodar Krai overnight on July 6, the Russian media outlet Astra reported.

Ukraine has not officially commented on the reported strikes, and the Kyiv Independent could not independently verify the claims.

An air alert was sounded in the city for several hours, and air defense was active. The consequences of the attack are still being determined, according to Astra.

The media outlet also published footage purportedly showing a burning maritime drone that was allegedly shot down during the attack.

Krasnodar Krai is located east of Crimea, with the Kerch Strait separating them at their closest point.

Ukraine regularly strikes military targets within Russia as Moscow continues to wage its war against Ukraine.

The Russian Defense Ministry claimed that Russian forces downed 120 drones overnight on July 6.

Thirty drones were shot down over Bryansk Oblast, 29 over Kursk Oblast, and 18 over Oryol Oblast, according to the ministry. An additional 17 and 13 drones were reportedly intercepted over Belgorod and Tula oblasts, respectively, the ministry said.

Due to drone attacks in Russia, numerous flights were canceled or delayed at several airports, including Moscow's Sheremetyevo Airport, overnight between July 5 and July 6.

Ukrainian drone strike on Russian airfield hits bomb depot, aircraft
The airfield hosts Su-34, Su-35S, and Su-30SM jets that Russia regularly uses in air strikes against Ukraine, according to the Ukrainian military.
Drones reportedly attack Russia's Black Sea fleetThe Kyiv IndependentOlena Goncharova
Drones reportedly attack Russia's Black Sea fleet
  • ✇The Kyiv Independent
  • Ukrainian-founded Grammarly to acquire AI email app Superhuman
    Grammarly, a company with Ukrainian roots, announced its intent to acquire AI email writing app Superhuman as part of its expansion into an AI productivity platform, the company said in a press release on July 1. Grammarly is the most valuable company with Ukrainian roots, reaching $13 billion valuation as of 2021. Grammarly was founded in 2009 in Kyiv by Oleksii Shevchenko, Maksym Lytvyn, and Dmytro Lider.According to Grammarly's press release, email is Grammarly's top use case, with the platfo
     

Ukrainian-founded Grammarly to acquire AI email app Superhuman

2 juillet 2025 à 07:45
Ukrainian-founded Grammarly to acquire AI email app Superhuman

Grammarly, a company with Ukrainian roots, announced its intent to acquire AI email writing app Superhuman as part of its expansion into an AI productivity platform, the company said in a press release on July 1.

Grammarly is the most valuable company with Ukrainian roots, reaching $13 billion valuation as of 2021. Grammarly was founded in 2009 in Kyiv by Oleksii Shevchenko, Maksym Lytvyn, and Dmytro Lider.

According to Grammarly's press release, email is Grammarly's top use case, with the platform editing over 50 million emails weekly.

Superhuman is an AI email application that the company says helps users respond to emails faster and reduces time spent on email communications.

Users are already sending and responding to 72% more emails per hour after using Superhuman compared to the previous period, according to Grammarly.

"This is the future we've been building toward since day one: AI that works where people work, not where companies want them to work," said Shishir Mehrotra, Grammarly's CEO.

The acquisition follows Grammarly's recent purchase of Coda, a productivity tool company. The combined platforms will allow users to work with multiple AI agents for different tasks within email communications.

Grammarly says that its service is used daily by over 40 million users, generating annual revenue of more than $700 million for the company.

As Russia ramps up missile attacks, US halts promised air defense shipments to Ukraine
Among the items being held back from Ukraine are Patriot air defense missiles, precision artillery rounds, Hellfire missiles, and drones, three sources told Politico.
Ukrainian-founded Grammarly to acquire AI email app SuperhumanThe Kyiv IndependentAbbey Fenbert
Ukrainian-founded Grammarly to acquire AI email app Superhuman
❌