There’s a possibility if Coda did as you suggest, they might not be around for the long run. Is that risk something you are comfortable with?
2 Likes
I 100% agree with the comments above about pricing concerns. My team and I switched away from Airtable because of their constantly constricting features on the free plan and the insanely high pricing once per-seat is factored in.
We’ve done so much work to build our Coda docs and I’m worried Grammarly is going to come in and mess everything up.
As a writer and editor, I’ve never been very impressed with Grammarly. It’s clunky, overbearing, and intrusive, whereas Coda is flexible, dynamic, and a breath of fresh air.
I hope Grammarly doesn’t change Coda. Instead, Grammarly should learn a thing or two (or a dozen) from Coda.
10 Likes
This reads like an ad and its wording is very similar to some of the materials on Grammarly’s own blog. May I ask if you work for them or are affiliated with them?
The post mentioned technical writing and documentation as a use case. I know many writers including myself who don’t want “popups” from AI cropping up everywhere while they write. There’s actually a trend towards minimalistic, stripped-back writing software at the moment. Is Coda aware of this trend?
This also caught my eye: “Increasingly, knowledge workers (like me) use the Grammarly Desktop, which places Grammarly in EVERY app I use, including–but not limited to–Slack, email apps, note-taking apps, database apps, design apps, and every imaginable place where I write, capture, and share information”
I don’t care how well-meaning Grammarly claims to be or how many benevolent promises it makes about safeguarding user data. I and many others don’t want any company or its AI being able to scan everything I write in every app and place on my desktop. It’s not normal, it’s annoying, and it sets a harmful precedent for the future of tech.
2 Likes
I agree. The announcement and the rationale seem 100% focused on AI, and that’s not what has made Coda useful and unique to me. I hope this merger does give Coda more financial resources to work on the features and fixes that we want. I would love to see the mobile app be much more useful, for example.
2 Likes
There is a lot of complaints about Coda docs being integrated with AI.
AI infused documents is the new normal. It doesn’t matter whether it is word docs, spreadsheets PowerPoint, or any of the new no code tools.
@Bill_French2 shared a link to a Microsoft presentation on their integration of Co-pilot with Office, I strongly recommend that you watch that video. Can
Let me illustrate with a yarn. For all of human history before the 20th century, written knowledge was shared on an inert surface by a human. In the 20th century we finally developed the ability to digitally store, manipulate and transmit knowledge to another person(s).
This lead to an explosion of information and knowledge, to the extent that knowledge workers spend as much as 40% of their time looking for information buried in files and folders.
Gradually, the writing surface gained limited intelligence, correcting spelling mistakes, and later made grammar suggestions. For the first time, the writing surface “talked back”.
Recently nocode tools started providing the ability to create smart documents with links and backlinks, easily accessible tables. Now information could be easily searched in corporate wide hubs and other ways of storing data.
This accelerated again with the advent of AI in its various forms. In the Microsoft video I mentioned above, they provide really amazing examples of AI meaningfully assisting in the creation of new knowledge. It can be slides prepared from text, analysis of spreadsheet data, anything you can imagine.
The only other platform in the same league as Microsoft, is the Coda/Grammarly/Snowflake combination.
Knowledge workers are now in possession of a writing surface that will with a Coda =, @, / AI prompt or a Grammarly button be able to engage in an intelligent conversation with you. It will search and refine.
Context aware of what you are writing, aware of relevant information in general, and corporate information specifically. While at the same time, it will also manage your permissions to access data.
No software developer of office type tools is going to develop something that does not include AI, it’s just not going to happen, indeed, it is going to accelerate.
And Coda/Grammarly had just moved up a gear. Or two.
2 Likes
I’ve never read the Grammarly blog, and I’m sure there are a few people on this forum who can say I am not an employee or paid shill for either company. I’m just an old dude who uses their products and achieve hyper productivity. My team does as well with Coda Enterprise edition.
Sure. I don’t work for them and while every Coda user is an affiliate by definition, I think I’ve earned $9.00 in referral credits since 2019. Perhaps the Coda folks can confirm this.
This is a common sentiment for writers. However, most of Grammarly’s customers are not writers. They’re information workers who write a lot. I think this is an important nuance required to understand the nature of knowledge work. To these workers, the connective tissue between their current focus of work and other relevant information is best implemented with proxy gateways. They’re ready when workers need them.
I can’t speak for the Codans, but I think they understand the minimalistic movement. I used Coda to create InFlight - a minimalistic writing canvas that simulates being locked in a tube at 40,000 feet. If you collapse the nav bar, it’s just you and the canvas.
It’s rational to fear the use of AI in everything we do. Even the most well-intentioned products do nefarious things. However, you are free to reject these ideas and I’m free to embrace them. While there’s a non-zero probability of a breach of trust, the risk is low and most companies have proven they can be trusted concerning AI and insulation of your data.
It’s kind of funny though. Your data is being purposed every day in ways that are far more nefarious than the worst AI companies. Media’s preoccupation with AI risk has reached irrational levels. There’s little evidence that AI companies are stealing private information for training unless they are granted such rights. In contrast, data breaches occur every day for information entrusted to systems that have nothing to do with AI.
By definition, innovation change is not supposed to be “normal”. It’s a new and innovative use pairing intended to benefit [future] customers in several ways. You are free to reject it and tell others why their making a mistake. I’m free to do precisely the opposite. You are probably a future former customer.
Coda has probably reached a nexus where a sizable number of customers will flee to other products that aren’t taking a big step and the associated risks of alienating dedicated, early customers. This is the normal cycle of products that change over time to anticipate important tech trends. Another typical cycle is product stagnation. Coda has chosen to avoid near-certain demise by pairing its tech with another successful company with a very similar vision but deeply complementary.
2 Likes
This is a wonderful reflective journey that helps us understand transitions. @Piet_Strydom makes it obvious that if you’re disenchanted with Coda’s new direction, you don’t fully understand the natural transformation of human knowledge.
Great perspective, Piet.
2 Likes
I have long wondered why Coda AI is completely useless for formulating formulas for your own application. Almost a bit embarrassing how much it hallucinates and is incapable. It would be easy to add the Coda formula language to the system prompt.
4 Likes
I don’t want to sound Ludite, but it’s already scary how much, let’s say Google, knows about us. Having a single AI tool be aware of every single interaction we have in the digital world (and probably at some point also in the real world, by having your phone/smart glasses stream all audio/video to it) will grant you super human productivity, but at what cost? What kind of risks are we exposing ourselves to?
The incentives for companies have not changed because of the raise of AI, so I see no reason why they will be more benevolent with the usage of our data than in the past. And just because major data breaches have not happened yet in any major AI tool, that doesn’t mean they won’t happen in the future. With so much context from our lives the consequences would be many orders of magnitude worse.
We have to be careful what we optimize for. If we only want to maximize productivity, at some point humans with our poor capabilities will only get in the way of advanced AI agents.
1 Like
Not optimistic about this at all.
We’ve integrated Coda into a lot of processes in our company based on the pricing and roadmap. A year ago they changed the pricing so it complicated or outright made many use cases impossible. Now they are doubling down on AI, and seem to be chasing a different market/use case.
What companies need is security to know that the time and money invested in a tool will suddenly not become obsolete and leave them scrambling to find different solutions with a few weeks notice (like with the previous pricing change). This announcement briefly touched on this, but didn’t reassure me enough that I can recommend sticking to Coda for more than the basic use cases which are covered by other tools. Certainly not for critical processes.
I am a big proponent of Coda in our company and steer everyone away from other solutions, but after this I will be looking into alternatives and planning on how to offboard in case the future execution goes as this announcement suggests.
4 Likes
This is simple to understand.
There are about 200 billion lines of solid open-source code from which to train LLMs about Python. There are near zero lines of Coda’s formula language in open circulation. Many unfamiliar with the data necessary to train LLMs assume it’s a simple matter of providing the CFL documentation, and the LLM magically understands how to use it in millions of cases. It’s not that simple.
Ergo, you should use AI for that. 
Its incompetence with CFL is a direct function of the vacuum of data about CFL.
Northing prevents you from trying this. Use Google AI Studio to create a Custom GPT. Reference the Coda CFL documentation, and then begin asking it to generate answers. See how well that works.
What Coda is doing now resonates with ideas I explored more than 30 years ago with the FrameWork IV application. The concept evolved through Microsoft’s OLE and DDE, spell check, and grammar correction, to what we have today.
That does not mean that I am paid, or that I should be paid for my rambles on the community.
I find that being positive, and embracing what the future holds, to be reward enough.
3 Likes
Yeah, many people find this a scary thought. But what you are thinking is not necessarily reality. Let’s examine the way AI works in the context of Coda AI.
You’re cursor is sitting on the canvas of a Coda document. You highlight a passage you’ve just written because you think it could be expanded and conveyed better to your readers. The highlighted the passage is copied when you click the AI icon and select “Elaborate”.
It is instantly sent to OpenAI over HTTPs using a REST API call. That API integration [likely] shares your highlighted snippet with an LLM to return a text completion. This is where everyone freaks out.
The interchange with OpenAI is (by default) an opt-out event because ALL API calls to LLM providers are (as a security standard) exempt from using your data for training purposes.
The resulting output is also exempt from use by virtue of their security policy which is assured across downstream integrators. Coda receives the resulting text completion output, and allows you to insert it to replace the highlighted passage.
Throughout this process, your data is anonymized. OpenAI has no idea or indication this passages is connected to you.
A lot could go wrong in an architecture like this, but a lot could go wrong at Coda. You’ve already entrusted everything–not just this passage–to Coda. It is a SaaS platform as is OpenAI. However, everything you write and all your confidential company data is already exposed to your SaaS providers. Exposing it to a downstream AI processes doesn’t change anything except you’ve added one additional attack surface.
Increasing attack surfaces is bad. However, it’s likely your data already has several and you’ve created these exposures in exchange for business benefits.
Grammarly works the same way, except, it can perform this process everywhere your cursor blinks - every app, database field, mobile form field, etc. However, the beauty of Grammarly is that it doesn’t increment the number of attack surfaces for all the apps where you use it. It’s a single point of possible attack.
That said, Coda is smart to eliminate security threats by aggregating risks to a single provider. This is a byproduct of the merger, but it is extremely advantageous because Grammarly is a trusted provider with a very good multi-decade track record for forty million users.
You can fret and worry about AI risks for the rest of your life, but largely, it’s probably a waste of time. It’s certainly a rejection of opportunities for potential business advantages if you choose to reject AI because of fear that your data might escape your control. That ship sailed long before AI became a thing.
3 Likes
Alright, let’s break it down:
- Coda’s been letting the community down for a couple of years now.
- The main issues we raised? Never really got addressed.
- “AI” is just a buzzword. I need AI for two things: a) importing data from audio, video, or text into structured format, and b) analyzing that data to provide insights. That’s it.*
- Shishir seemed to be looking for a way out of the mess.
- There’s never been a clear roadmap showing real solutions were coming.
- Looks like Shishir found an exit for himself (don’t fool yourself - this is business exit, it’s not an upgrade; I’ve done similar too).
- Only maybe he even found a temporary solution for the existing Coda docs.
- This announcement? Nobody really understood it, and it didn’t feel like the reassurance we needed.
- If Coda actually had a solid future, they’d have shared it by now.
- @shishir and Codans are the only people who can stop the hatred and distrust by being open and honest
At the end of the day, everyone’s got to decide for themselves: keep hoping or move on. It all depends on what you’ve got at stake.
IMHO
7 Likes
*All other things related to LLMs (let’s not exaggerate with “AI”) like summarization, create tables, generate images, agents, assistants, etc. is already happening with browser add-ons and next Chromium and the OpenAI browser will have them by default. I don’t need resource-burning, buggy, brain-destroyer like Grammarly (Yes, I used it for years as non-native. And I come from the localization industry so we pushed it to the limits. Grammarly is only good in English and copywriters thinks it sucks)
1 Like
Really naivé question - what does Coda have to do with Grammarly?
I don’t understand how they can merge into one cohensive application.
Hard to imagine Coda docs being written by AI.
1 Like
These are each valid points to consider, and you are right - it’s an exercise of reflection on the present and future direction you believe the product is heading. But, it’s also possible that you picked the wrong horse (for your interests) long before this merger.
Coda is not impervious to economic realities, and as skilled technologists, we each have a duty to ourselves and our constituents to make rational judgment calls concerning the probability of sustainable platforms we build upon.
I was under no illusion that @shishir or Coda any of their talented Codan’s owed me anything. I paid for Enterprise and they delivered precisely my expectations. Had they failed to do that, I would have stopped paying.
If you think there’s a greener pasture, you need to hop the fence and when you do, please share with us the GPS coordinates of that pasture.
This assertion holds no sway with me and likely the 40 million Grammarly users because vastly, knowledge work is not performed by copyrighters. Arsonists think firetrucks suck. What’s your point?
@Richard_Kaplan
Please see this post that I made to a similar question.
Isn’t it funny how, for such an obvious opportunity, there are always people insisting it’s not because, technically, you could piece it together using extra tools and platforms? It’s not that I don’t know how to build my own Coda AI / GPT / systempropt for using formulas. The issue is the missed use case. I’d love to promote Coda to more users and if they could simply use the integrated AI for building, they’d be far more likely to adopt it. Telling them to go DIY their own custom Coda AI doesn’t provide value to them and neither for Coda.
3 Likes