[HN Gopher] Canvas
___________________________________________________________________
Canvas
Author : t55
Score : 135 points
Date : 2024-12-10 18:01 UTC (4 hours ago)
(HTM) web link (openai.com)
(TXT) w3m dump (openai.com)
| t55 wrote:
| 1. Available for Everyone 2. Python Code 3. Custom GPTs
| rawgabbit wrote:
| TLDR: Canvas is a new side by side mode in ChatGPT.
|
| You can use it for improving documents. You interact with ChatGPT
| on the left panel. You see the document on the right panel
| including the editorial comments by ChatGPT.
|
| You can also use it for running and debugging Python code. You
| interact with ChatGPT on the left panel. You see the code and the
| console messages/errors/output on the right.
|
| Demo video url: https://www.youtube.com/watch?v=qZ0ImE41pVs
| zingababba wrote:
| Sounds like the Claude artifact panel?
| aaronharnly wrote:
| Yes.
| virtualcharles wrote:
| It's very similar, but the big difference is that canvas lets
| you select a specific line/paragraph/word/etc and "talk" to
| just that selection.
|
| On the flipside, it only shows formatted text, it won't
| render anything if you're working with code.
| scosman wrote:
| I've been using Cursor for these writing use cases for a
| while (markdown, not code). Select part, ask for edits.
| It's quite handy. This actually sounds useful.
| binkHN wrote:
| > You can also use it for running and debugging Python code.
|
| Not limited to Python code.
| chipgap98 wrote:
| It can only _run_ python code, but it can write other
| languages
| llm_nerd wrote:
| Though note that it runs python code via web assembly, on
| the client's device, using pyodide. Pretty easy way to
| avoid DOS attacks or even any resource usage on their
| servers.
|
| Ostensibly they could add other platforms/languages that
| have wasm implementations.
| KTibow wrote:
| I wonder if the data analysis tool will switch to using
| the same backend in the future
| NullHypothesist wrote:
| Does it work for o1 or just 4o?
| llm_nerd wrote:
| Just 4o.
| airstrike wrote:
| Yeah, that's the big question. Wake me up when it works with o1
| infecto wrote:
| Will find out once the replay is up but it would be nice if
| someone noted the actual changes from the current state of the
| beta version.
| submeta wrote:
| What exactly is this? I can't find any info in the link. It just
| says ,,replay coming soon."
| scosman wrote:
| Had same question. Writing+coding collaboration tool
|
| Better link: https://openai.com/index/introducing-canvas/
| odirf wrote:
| Already posted.
| kristopolous wrote:
| although these things are impressive at writing hello world
| in a bunch of ways, I've really found them pretty useless
| inside any sufficiently complex codebase. Am I just
| incompetently using them?
| benbayard wrote:
| I use claude every day to help me write code. I doubt it's
| incompetence, but maybe a lack of trust. Cursor + Claude
| does great and I use it to do real work for my business.
| criley2 wrote:
| Using the front end of these websites to work in your
| complex codebase is very challenging. I use the chatbots
| for higher level questions about libraries and
| integrations, not for specific implementation details in my
| codebase. But without a data agreement in place, you
| shouldn't (or maybe can't) paste in code, and even if you
| could, it's an inferior way of providing context in
| comparison to better tools.
|
| However, I do use copilot+vscode with claude 3.5 and the
| "Edit with Copilot feature" where it takes my open files,
| plus any other context I want to give it, to drive changes
| to my files, has been surprisingly good. It's not really a
| time saver in that the amount of time I spend verifying and
| fixing or enhancing the result isn't really faster than me
| just writing it myself, but I still find benefits for
| brainstorming, quickly iterating on alternate ideas and
| smaller refactors, and overcoming the "get started"
| hesitation on a seemingly complex change. I'm to the point
| where it can absolutely add tests to files using my
| existing patterns that are well done and rarely need
| feedback from me. I have been surprised to the see the
| progress because for most of the history of LLM I didn't
| find it useful.
|
| It also helps that I work in a nodejs/react/etc codebase
| where the models have a ton of information and examples to
| work with.
| rtsil wrote:
| > But without a data agreement in place, you shouldn't
| (or maybe can't) paste in code
|
| There's a checkbox you can toggle so that Openai doesn't
| use your code to train their models.
|
| And I find the "chatbot" experience different and better
| than aider/copilot. It forces me to refocus on the really
| useful interfaces instead of just sending everything, and
| makes it better to verify everything instead of just
| accepting a bunch of changes that might even be correct,
| but not what I exactly want. For me, the time spent
| verifying is actually a bonus, because I read faster than
| I can type. I think of it as a peer programmer who just
| happens to be able to type much, much faster and doesn't
| mind writing unit tests or rewriting the same thing over
| and over.
| criley2 wrote:
| The problem with reading vs writing is building "true
| understanding". If you are reading code at a high level
| and building a perfect mental model and reading every bit
| of code, then you're doing it right. But many folks see
| finished code and get "LGTM brain" and don't necessarily
| fully think out every command on every line, leading to
| poor understanding of code. This is a huge drawback in
| LLM-assisted coding. Folks re-read code they "wrote" and
| have no memory of it at all.
|
| In the edit experience I am using, the LLM provides a git
| style changelog where I can easily compare before/after
| with a really detailed diff. I find that much more useful
| than giant "blobs" of code where minor differences crop
| up that I don't notice.
|
| The other massive drawback to the out-of-codebase chatbot
| experience (and the Edit With Copilot experience IS a
| chatbot, it's just integrated into the editor, changes
| files with diffs, and has a UI for managing file context)
| is context. I can effortlessly load all my open files
| into the LLM context with 1 click. The out-of-editor
| chatbox requires either a totally custom LLM with various
| layers to handle your codebase context, or you have to
| manually paste in a lot of context. It's nonsense to
| waste time pasting proprietary code into OpenAI (with no
| business agreement other than a privacy policy and a
| checkbox) when I can get Copilot to sign a BA with strict
| rules about privacy and storage, and then 1-click add my
| open files to my context.
|
| Folks should give these new experiences a try. Having
| claude chatbot integrated into your editor with the
| ability to see and modify your open files in a
| collaborative chat experience is very nice.
| trolan wrote:
| I find it's directly proportional to the inputs you give
| it. So either you need to break down the requests into
| smaller boxes, or use a tool like Cursor or my preferred,
| Aider. They both automatically bring in more context about
| your codebase, such as Aider bringing in your whole git
| repo map, to help it integrate better.
| kristopolous wrote:
| my biggest problem in coding is when I'm looking at
| things spanning 10 directories and 100 files with totally
| useless name like controller/viewer.js,
| viewer/controller.js, viewercontrols.js, etc and try to
| find where, 15 function calls ago and 80 statements ago
| some bug occurred. I was hoping that AI could help
| untangle these messes.
| Maxatar wrote:
| Not at all incompetence, it's just a very open ended UI for
| a tool that is best used for specific use cases. It can be
| tempting to try to get it to solve very broad problems
| since the UI lets you do that, but where it excels is when
| you use it to save time by constraining it to do the grunt
| work for you.
|
| Don't ask it how to solve a problem for you, instead write
| out the solution in plain English, provide an API for it to
| fill in, and let it write out the solution with unit tests.
| miki123211 wrote:
| Canvas (and Claude Artifacts) aren't things you use inside
| a large codebase, they're more for one-off tools that you
| may want to run, without the hassle of setting up a stack
| for each, copying code between the LLM and your local
| editor etc.
|
| Think "here's a schema for some internal API, write me a
| tool that lets me put in a product ID and see the number of
| purchases, chargebacks and returns for that product over
| time, on graphs, also provide sortable by-country tables
| for these stats".
|
| With artifacts, software goes from something you make once
| and use for a long time, to something you make at need, to
| do exactly what you need and no more, and then immediately
| throw away once it has served its purpose.
| thekid314 wrote:
| Yeah, I think this is a shift that will be very
| interesting when commercialized. I already use chatGPT to
| write odd plugins for wordpress and programs for image
| and text manipulation. Now I can get an applet to do
| almost anything simple after an hour of messing around.
| tpowell wrote:
| +1 for Cursor.
|
| This guy [https://x.com/PrajwalTomar_] has been exploring
| workflows that involve using ChatGPT to assist in creating
| a Product Requirement Document (PRD), then using V0 by
| Vercel for mockups and bringing it all together using
| Cursor, maintaining continuity with markdown documents
| (.md) of the PRD and relevant database schemas etc inside
| the project to maintain continuity.
| dang wrote:
| Ok, we've changed to that URL from
| https://openai.com/12-days/?day=4 above. Thanks!
| yawnxyz wrote:
| probably a bug
| sergiotapia wrote:
| what is this (while the video isnt present)?
| andrewstuart wrote:
| I used the OpenAI collab tool yesterday for a couple of hours.
|
| I was trying to write a technical document.
|
| Very clunky user experience. The AI felt like an annoying
| collaborator who took control when I was trying to do my job and
| didn't obey when I asked it to do stuff and the basic markdown
| editing options were missing.
|
| Switched to another non AI editor and got the job done quick.
|
| Having said that, I like the idea and would use, but I do not
| want to ask AI to do find and replace. If they can flesh this out
| and improve it and make it "writer first, AI second" instead of
| "AI first" then it might be worth using.
| fosterfriends wrote:
| Looks like you can only use it with 4o at the moment, and not yet
| with o1 models. I might stick to o1 for editing jobs for the time
| being.
| fosterfriends wrote:
| Cool to see more attention getting paid to AI code review and not
| just code-gen:
|
| > Review code: ChatGPT provides inline suggestions to improve
| your code.
| wongarsu wrote:
| Feels a bit like we are going full circle: From an interface that
| contains a document that the AI will complete or modify to a chat
| interface, to an interface that contains a document that the AI
| will modify.
|
| Obviously this is a lot more refined than the rather simple
| playground interface of GPT3.5 that they eventually discontinued.
| But it feels a lot like going back to the roots, with the benefit
| of many of the lessons learned over the last two years.
| cloudking wrote:
| Canvas was releaased on October 3, 2024 - is this something
| different?
| asadalt wrote:
| i have had this for a while now (a month maybe?). I wish this had
| o1 option. nobody uses 4o for coding anymore.
| sunaookami wrote:
| I found o1 to be overly verbose and rather "meh" for coding.
| Tried both o1-preview and o1-mini, haven't tried the new o1. I
| loved using o1 to draft implementation ideas, things to note,
| etc and hand them off to GPT-4o to write them though.
| mellosouls wrote:
| Incorrect title. It's already released some time ago.
|
| Difference appears to be it's now available for free users.
|
| sama announcement here:
|
| https://x.com/sama/status/1866555731149045990
|
| _canvas is now available to all chatgpt users_
| ipsum2 wrote:
| That's also not correct, they added a bunch of new features to
| canvas, including a web based python interpreter.
| dang wrote:
| Recent and related:
|
| _Canvas is a new way to write and code with ChatGPT_ -
| https://news.ycombinator.com/item?id=41732634 - Oct 2024 (679
| comments)
| 1oooqooq wrote:
| what's up with everyone abusing canvas? i was confused enough
| when a drawing webapp used it and it didn't mean the canvas html
| element. now not-openAI uses it for something that isn't even
| related to painting.
| halyconWays wrote:
| What is this 12 Days of Christmas thing, with Sam Altman sitting
| there? Do they really think they're perceived as benefactors or
| relatable, lovable startup scamps? By anyone following the AI
| space they're the enemy. They've made a mockery of the entire
| purpose of their company, ousted the board that resisted their
| direction, are abusing the charity laws while making billion-
| dollar for-profit deals with Microsoft and defense contractors,
| and they abandoned every single promise towards openness. Even
| their name is a lie.
| butterlettuce wrote:
| I don't get the hate for OpenAI. They've made life easier for
| me and I'm not going to pretend they haven't.
|
| They came out with a revolutionary chatbot that provides actual
| value. A few months later, every single company starts pushing
| their AI offerings (Gemini, Apple Intelligence, Grok, AI this
| and AI that). Sam and Co needs to compete.
|
| > abusing the charity laws while making billion-dollar for-
| profit deals with Microsoft and defense contractors.
|
| Who cares? Compute costs aren't going to pay for themselves and
| money doesn't grow on trees.
|
| If they become ClosedAI, so be it. I'm still buying their
| stuff.
| halyconWays wrote:
| >Who cares?
|
| Who cares that a 501(c)(3) formed a shadow corp that's making
| multi-billion dollar deals with Microsoft and defense
| contractors, put a former NSA chief on its board, has stopped
| releasing all weights, and is scraping the entire Internet
| (and probably all your PII) to train its models in direct
| contravention of their mission statement to only do things
| that benefit humanity? Well, intelligent people care about
| that. The question you should be asking yourself is why don't
| you care?
| wiremine wrote:
| Looks like it's still in beta?
|
| I tried it out a few weeks ago and it was extremely buggy. I'm
| curious how it has been for others.
| virtualcharles wrote:
| I had the beta as well and have been using it quite a bit. I
| like the concept a lot. This is exactly how I want to be using
| AI for writing purposes. Overall, it's whatever content I want
| to write in, but then I can ask it, "Hey, help me improve this
| line," or "Fix this paragraph," piecemeal.
|
| That said, it only worked for me maybe 70% of the time. It
| would either have trouble finding things, or it would just
| arbitrarily rewrite the entire document, even when it was
| supposed to be focused on a certain part of it.
| tiahura wrote:
| So, basically everything but 4.5.
| paul7986 wrote:
| Anyone else want an AI phone that would work and it's interface
| would be like chatGPT's iPhone app? A stark interface with some
| app icons like current phone UI.
|
| Using chatGPT's iPhone app I am able to have an entire back n
| forth full conversation with it and able to get things done. Soon
| GPT will have vision AI and can discuss/work with what it sees or
| I show it. Apple Intelligence and Siri won't offer what I discuss
| here til Spring 2026. That gives competitors time to create and
| release an awesome new phone experience!
|
| I'd drop iPhone for a GPT (Microsoft) phone/ device quickly!
|
| *Slightly off topic but hoping it's one of Open AI's
| announcements ;-)
| raincole wrote:
| It looks nice and I might try it out. However I kinda hope OpenAI
| to stay as an API provider and let third parties do the UI.
| meta_x_ai wrote:
| OpenAI has completely pivoted to a Product company (vs a Model
| / API company)
|
| The minute they switch to API is the time their userbase
| realizes that OpenAI neither has the best model nor the most
| cost-effective one nor the fastest one.
|
| So, their business strategy is subscription and bundle a bunch
| of products and market it to the mass.
|
| They have a brand "chatGPT" recognition and they'll milk that.
|
| Sam is more of a Steve Jobs than a Bill Gates personality or at
| least wannabe
| Sateeshm wrote:
| The full pivot to products is another tell that they don't
| have high hopes on AGI
| bubaumba wrote:
| You may think Anthropic is different? It's the same UI + API.
| The difference is they don't create free accounts for
| personal emails. Sort of shadow banning. However they
| accepted my work email for free account and personal email
| for payed API. May be that's just me, I tried several
| personals.
|
| As for model's quality they are both impressive. I haven't
| run meaningful comparison tests.
| theshackleford wrote:
| > May be that's just me
|
| I would suggest this may be the case.
|
| > The difference is they don't create free accounts for
| personal emails.
|
| I'm using a personal account for free just fine, as are
| many in my circle.
| raincole wrote:
| I don't know what you imply by juxtoposing Jobs vs Gates.
| From what I know, Apple successfully marketed their hardware
| as a lifestyle. I don't think OpenAI is anything like that.
| meta_x_ai wrote:
| chatGPT's popularity and almost becoming a verb, says
| otherwise
| bubaumba wrote:
| They are both. and there are UIs for their API, which costs $$.
| But web UI is free. And limited. It's nice to have options.
| pkkkzip wrote:
| these announcements from OpenAI no longer has that exciting
| impact because of the repeated lies and disappointments in actual
| real world usage.
|
| my attention instead is focused entirely on open source models
| and its competitors which are already pulling ahead without the
| need for fancy announcements.
|
| Lastly, I really despise Sam for trying to cash out his non-
| profit RSUs and Elon Musk will most certainly prevent that from
| happening. While I'm not a grok simp or pay for X, I can already
| see OpenAI losing relevance.
|
| For example announcement from Deepmind vs Sam's expressionless
| interviews has different energy.
| tptacek wrote:
| What repeated lies and disappointments are you talking about?
| saurik wrote:
| FWIW, they lost my belief in them with the "advanced voice"
| debacle: they seem to have announced it when they did merely
| to scoop Google, not because that was a reasonable time to
| announce it :/. It then took them forever to _actually_
| release it, and while I totally use it (a lot: I am not
| saying it is a categorical failure or anything), it isn 't
| _anywhere near_ as good as their demo made it out to be; the
| result is that I simply don 't believe any of their demos or
| announcements now: I'll believe it when I see it and evaluate
| it on its merits only then... watching their videos or
| reading their press releases is probably a waste of time and
| energy.
| tptacek wrote:
| Got it. I think my bias in this is that 4o is my go-to chat
| model, and chat models are basically all I use; if you're
| doing more advanced stuff than that, I've got no
| visibility, and am inclined to just believe everything
| people say.
| _giorgio_ wrote:
| OpenAI is having on my life the biggest impact. More than
| Microsoft, Google, Amazon. It having a bigger boost on my life
| than my engineering degree.
|
| It's simply incredibly good.
|
| I couldn't care less about no profit.
| amelius wrote:
| Somewhat offtopic, but I noticed that ChatGPT in a Firefox tab
| makes Firefox unstable. Often when Firefox misbehaves, I close
| the ChatGPT tab and things go back to normal. Did anyone else
| notice this? I'm running Firefox on Ubuntu 22.04, 64bit Intel,
| and I have plenty of RAM (64GB).
| wingmanjd wrote:
| I haven't had the same experience.
|
| PopOS, 22.04 Firefox Also 64GB of total RAM.
| ericra wrote:
| No problems here with a very similar setup (22.04, Firefox),
| and this is on a 2015 laptop. Very stable and uses very little
| background resources.
|
| I am running unlock origin and privacy badger, but they don't
| block much of anything there.
| amelius wrote:
| Ok, thanks, maybe it is time for me to run a memtest.
| _giorgio_ wrote:
| Same config, no problem.
| serjester wrote:
| Regardless of all the internal drama, you've gotta hand it to
| their product team - they know how to ship. Personally I find
| their canvas tool best in class and some of these new features
| seem absolutely awesome. WASM python? This opens up so many
| possibilities and it seems likely they're building towards
| Bolt.new competitor.
|
| On that vein, it seems like every time OpenAI launches something
| it becomes table stakes for all the other players. I have no idea
| how Mixtral or any of the smaller companies will compete long
| term.
| cainxinth wrote:
| They have also clearly taken the "we have no moat, and neither
| does OpenAI" quote to heart.
|
| They know that the UX wrapper is likely going to be their only
| differentiator and want to build the walls of their garden as
| fast as possible.
| aylmao wrote:
| This could be tricky for them, because just as OpenAI has
| smaller competitors, it also has bigger ones.
|
| If Google/Microsoft adds these features to Docs/Word-- why
| would you use Canvas to write documents instead of the proven
| office software you and everyone you work with already uses?
| The biggest differentiator for OpenAI becomes again in this
| case the quality of the AI.
|
| EDIT: on this note, Apple's approach of commoditizing AI
| writing tools to the OS-level is also interesting. If AI
| writing tools are available on every text-box, there's no new
| product in building a new AI-powered text-editor.
| SubiculumCode wrote:
| My issue with Canvas is that its not available with o1, and if
| you are working on a large function it can just truncates the
| prior script there.
| Maxatar wrote:
| If you find it truncating output, enter the word "continue".
| HaZeust wrote:
| I don't think they do know how to ship. Reaching this feature
| is at the discretion of the model and how it prefers to answer
| the question. If I want to use canvas for a coding-
| brainstorming session and the model doesn't think the prompt is
| rigid enough to invite it as a use case; I'm SOL.
| hipadev23 wrote:
| https://openai.com/index/introducing-canvas/
|
| Maybe I'm old fashioned but I loathe video updates without a
| companion text version.
| theshackleford wrote:
| Whilst i'm a little ashamed to share this, it does seem relevant.
| I've been using canvas for a while and have found it valuable for
| my needs.
|
| Due to medical reasons, I'm on daily medications that heavily
| impact my cognitive abilities. This is combined with chronic
| pain, which affects every aspect of my life. Thinking, energy,
| stress etc. This has made everything harder. In fact it was
| suspected/suggested at one point, that I would never return to
| the workforce, though luckily this did not come to pass.
|
| While I still retain my knowledge and can learn new skills, brain
| fog and exhaustion from my physical symptoms have made it
| difficult to communicate as effectively as I once did, which is
| particularly frustrating since communication was prior to these
| issues, my biggest strength. Now, I often struggle to express my
| ideas in a clear, digestible manner. It's very difficult to
| gather all my thoughts in the one place coherently for complex
| topics.
|
| To assist with this, I use ChatGPT with Canvas as I guess a
| structured editing tool. I input my content and research,
| organise it to the best of my ability, and ask for feedback to
| refine the presentation. Clarifying structure, reducing
| redundancy, and improving flow. I don't ask or let it insert data
| about the topic itself, I ask only that it modify how my data is
| communicated.
|
| It's a bittersweet solution, but I would say that it's helped me
| remain employed, which is important, because I have a lot of
| ongoing medical costs.
| SuaveSteve wrote:
| May someone please say how this compares to Claude's project
| view?
___________________________________________________________________
(page generated 2024-12-10 23:01 UTC)