[HN Gopher] Launch HN: Dart (YC W22) - Project management with a...
___________________________________________________________________
Launch HN: Dart (YC W22) - Project management with automatic report
generation
Hi HN, We're Zack and Milad and we're building Dart
(https://itsdart.com). Dart is a fully-featured project management
tool that can automatically create reports, fill in task
properties, generate subtasks, detect duplicate tasks, and more.
Here's a quick demo of some of the highlight features:
https://youtu.be/CMsBAv9CCyU Like many others we grew frustrated
with the constant upkeep in creating and organizing tickets in Jira
and other PM tools. We routinely wasted as much as seven hours a
week on repetitive PM chores like cleaning up the backlog or
drafting changelog updates. At Zack's previous company he even
built a system to try to automate some of this work. In
conversations with other founders and engineers we learned that he
was far from the only person doing this. We started Dart when we
realized we could bring a new approach to this problem through
techniques enabled by generative AI. There are of course a lot of
project management tools out there, but none have the ability to
automate most of the busywork without having to configure long sets
of rules. For example, one of the most helpful functions is
summarizing work that was accomplished over some time interval, by
automatically creating a changelog update for it that can be pasted
into company blogs or elsewhere. Another highly used feature is
giving Dart a large task or PRD and generating a recommendation for
how to break it up into more manageable subtasks. Over the last
two years we've continued to build a broad set of PM functionality
in Dart (task management, docs, roadmaps, etc.) along with what we
hope is a neatly vertically integrated application of gen AI that
adds value. Dart is built with Vue on the frontend, Django on the
backend, and uses various gen AI models such as GPT-4. We embed
almost all of the content that goes into the app for later
retrieval and search. Over time we build up context for users,
workspaces, working patterns, etc. and are able to provide this
context to do situation specific few-shot prompting or RAG,
depending on the situation. For better or worse we've implemented
most of our LLM infrastructure internally and are largely calling
OpenAI APIs directly. While Dart improves its recommendations as
users work on more tasks, ultimately it's not quite ready to
replace your product manager. You'll probably still need to
evaluate if the recommendations make sense for your team, and make
any adjustments before confirming them. By having a helpful
companion offering suggestions as you go we've still found that
this process ends up net saving time-about 12% of time spent in PM
tools according to our users. You can think of it like a helpful
tool for generating ideas as opposed to a product that will fully
strategize and do all of your work for you (at least for now!).
Our current users come from a wide range of backgrounds including
other founders, developers, writers, small business owners,
consultants, researchers, etc. You can try out Dart for free at
https://app.itsdart.com/sign-up. We'd really appreciate any and all
feedback or ideas in the comments!
Author : miladmalek
Score : 61 points
Date : 2024-02-20 15:24 UTC (7 hours ago)
| leoqa wrote:
| I went in pessimistic but the demo was pretty good- every AI-
| based feature was solving a real point of friction and seemed
| reasonable.
| zswaff wrote:
| Thanks! At this point AI is obviously a bit of a buzzword, so
| we've paid extra attention to only applying it when it actually
| adds value
| TheFragenTaken wrote:
| Even if I am probably not going to use it (too heavily/deeply
| invested in Azure Devops), I just wanted to give kudos to your
| product demo. By itself it's straight to the point, doesn't
| ramble, and calmly highlights whats important about your product.
| Compared to your other videos on the YouTube channel, it's a
| dramatic improvement. Congrats on the launch.
| miladmalek wrote:
| Thanks! Yeah going through the process of making this demo
| efficient was definitely a good exercise. It's always harder
| than I think it will be as my default is admittedly kind of
| rambly.
| samstave wrote:
| This would be great:
|
| * Resource loading, availability, tasks summary, personal task
| completion reports..
|
| * Maintain history of task time, cost, resource, as assigned by
| tasks, and resources completing their tasks.
|
| * Templates of tasks, or projects where if you say have completed
| a common project that have tasks you do (as if your a consultant
| with various clients with similar projects you can pull a gant,
| resource load, historic times as a template - then use them
| across your Program - and have a report on how similar task
| groups etc across your program vary.
|
| * Typical completion resources required so that if you simply say
| the name of a task, it will say "To do this, it usually requires
| ABC - but right now you only have AB available - so your costs
| and resource availability for this task on this project is X
| based on current loadout to resources.
|
| * Track what external resources were needed in historic tasks "To
| do this - you'll need these external resources, here is a summary
| email proposal to that external consultant to get costing and
| availability - effectively using your project plans to generate
| mini RFPs to the subs/trades/disciplines/departments needed on
| that task and have it make it easier for not only them to
| responde, but when they do - have it automatically give you the
| diff impact to the project/program/task/schedle/cost....
|
| * as you build out project portfolio, you just throw bullets in
| and it will pull from previous projects and put in the "project-
| lets" and allow you to build them in a more automated fashion.
|
| * in construction/deployment/commissioning/go-live/acceptance
| testing (in DCs and Health, and tech), if you can give it a
| primavera/ms proj/whatever - even a PDF - and have it read what
| some EXTERNAL PM assigned to you as a sub on the project and have
| it adjust your section based on your available inputs and
| effectively import your tasks from some external PM
|
| * have it provide your company with a really well written
| portfolio completed projects.
|
| * self-service reporting capabilities to external stakeholders...
| ESPECIALLY CFOs.... with a little dashboard widget that you can
| just point people to "dart.MYCOMPANY.COM" which could be a
| status-page like dash for whomever you allow to see highlevel
| report on demand for your
| program/projects/departments/budget/etc...
|
| * give scheduling reminders for upcoming resource requirements. -
| like status checking delivery dates from vendors with simple
| automated requests for "is component X still scheduled for
| delivery on DATE" - conversely - sending alerts to delays that
| will affect subs/trades who need access to a site/area whatever
| (like concrete guys are two weeks behind, so your delivery area
| for materials is affected, mr steel guy...)
|
| EDIT: I forgot about the BANE of any large project: CHANGE
| ORDERS.
|
| Usually these come from project owner/designer -and can have
| _significant_ impacts on a project sched /cost/resource avail...
|
| If you can track and manage change orders, construction
| manages/TPMs/any stakeholders will find value in it enough if you
| can show a report on the history of the projects' intent
| (budget/end-date) especially because change-orders is usually a
| two-way street on the negotiations btwn a subs/owner on who pays
| for what in change order...
|
| example: arch designs "thing" - sub implements it. now three
| scenarios happen:
|
| 1. All good.
|
| 2. Arch/owner F'd up intent to sub, sub has to fix at certain
| costs (the owner will say "your job to know my intent" -
| subs/trades say "nope, show me the money")
|
| 3. sub f-d up and they have to eat all/some portion of cost.
|
| REDUCING CHANGE ORDERS through such comms that I mention above is
| a golden goose in and of itself - but only you can prevent
| dumpster fires!.
|
| Finally it ties back to portfolio "reduced change orders by %
| across N% of our portfolio etc...)
|
| (change orders from a sub, obv comes out of their fees/profit -
| so change order mtgs on big ticket issues are REALLY fn painful
| mtgs because its a "who communicated what where when with whom"
| battles...
|
| ====
|
| Attempting to build a project based on this comment:
|
| https://i.imgur.com/ZYsTSk2.png -- this is weird, it just adds
| checks at bottom of text input - should collapse orig input
| prompt.
|
| I cant click anything - https://i.imgur.com/LXn2RVU.png
|
| Am I only supposed to have one task in this box?
| https://i.imgur.com/O3bY286.png
|
| OH I get it... you should have it auto suggest "extract" from
| each bullet... https://i.imgur.com/2yFXiqd.png -- the UX on this
| box is confusing...
|
| Or trigger a "would you like to extract/whatever - and have it
| cycle through each bullet. https://i.imgur.com/2yFXiqd.png
|
| Also - need tool tips. Why the stop-sign?
| https://i.imgur.com/Kp1FGI7.png
|
| "I'm posting too fast" - so Ill have to just write you guys a
| separate post later I guess - but the weird UX halts, no exp on
| what "/ for AI" means out the gate was clunky, but still dope.
|
| (might be better to have your system have an-onboarding example,
| where you do ALL THE THINGs - and ask the user to reject your
| examples - kinda like Scenario has prompt chunks with an X if you
| want to kill that action...)?
|
| What do from here: https://i.imgur.com/HhVU8vK.png - deleted all
| text, kept "tasks" and cant click anything but X? -- I clicked X
| - it asked me to DISCARD/KEEP Draft... I KEEP - Where drafts?
| https://i.imgur.com/1Pb4UMM.png
|
| I'll build a an email to you guys, from this....
| zswaff wrote:
| Wow, really appreciate the detailed list here.
|
| Some of the reporting, history, dashboards, templates, etc.
| that you mention is covered now but Dart probably isn't quite
| ready to solve some of your more advanced needs. It's all
| pretty aligned with our vision in the long run though. We're
| particularly excited about all of the ideas around resource
| tracking and allocation--we'll get there! Very curious to hear
| how it works for you.
|
| What's your use case for
|
| > a really well written portfolio completed projects
|
| ? Haven't heard this much before. Useful for an agency to
| showcase work?
|
| edit: formatting
| samstave wrote:
| I've built MILLIONs of SQ ft of hospitals, corporate, DCs,
| tech, commercial on billion+$ budgets etc... I've done
| AMAZING work.
|
| I never kept a good portfolio of my own accomplishments, and
| it would have really helped me as an independent consultant.
|
| I've seen so many smaller consultancies who are quite capable
| get "out marketed" by bigger consultancies - To have a
| running, ongoing portfolio - that you dont have to have
| another resource (pm/graphics/web/copy person... you can have
| a more compelling portfolio highlighting your success in the
| project AS YOU GO - so that when you choose to put a
| completed project on your /about-us - you arent doing
| anything from memory.... if resources or * _PMs*_ leave your
| org - they dont take the value of their contribution alone in
| their head so you cant highlight the success of that part of
| your company history from memory
|
| Departments can have their automatically weekly/monthly
| progress/success/issues/updates status reports which is an
| extremely aggregating and tedious thing to do - the easiest
| way to stress out a teams productivity is if they have to
| spend an inord amt of time attempting to craft a status
| report through having all sub PMs/leads/whomever also having
| to attempt to recall WTF they did...
|
| Allow a resource to give a voice update and have AI pull that
| into a salient status update.
|
| Have pictures taken of [issue] (punch lists) and have them
| described by the subs/whomever -- and then just say
| "electrical closet 4100 has no firestop, conduit is in wrong
| place" -- and it updates issues, reports, status and notifies
| whomever is required (the issue shows up on the LV subs todo
| with pics, a voice note - and all other associated reports
| are updated.
|
| It can even be used to create "launch/go-live announcements"
| -- When we were building out tons of Salesforce floors - we
| were commissioning then handing off each floor to each depts
| that occupied those floors as we moved forward. Then it keeps
| track of progress, but also track open issues and
| resolutions... so even after hand-off there are so many
| little punch items that need doing - and sometimes the corp
| IT/facilities guys have to deal with too many people to
| address/report/assign and track outstanding punches.
|
| This can also be used to track overall quality of work - and
| you can have a portfolio of subs across projects/sites/states
| and say which vendors are performant/in budget etc...
|
| Basically an LLM for your entire program.
|
| [These were feedback I basically gave PlanGrid (YC) when they
| launched... as I was already CFO/CIO deep in these issues...
| They got acquired by Autodesk.
|
| You should be acquired/integrate/compete with them...???
|
| Or, since they didnt follow my inputs - take me on as a temp
| PM and compete. :-)
|
| --
|
| I know that this is initially internal dev-project focus...
| but all these apply same to large scale construction
| programs.
|
| + portfolio and succesful project reports with successful
| metrics for the build could be an additional service fee that
| an agency could charge/include in their proposals
| :DELIVERABLES: X shall provide as included in this RFP a
| complete project timeline of services, outcomes and XYZ from
| project that OWNER may use in their own success stories...
|
| etc....
| zswaff wrote:
| Yep, that makes a ton of sense. I totally see the value in
| building a portfolio like that.
|
| Well, Dart can do a bit of the shorter-term weekly/monthly
| project reports/updates now, but I can definitely imagine
| adding 'full project summary for portfolio building' to the
| types of reports we can generate. Like you said, great for
| the team and a cool resource to hand off in deliverables if
| the owner wants to include as a success story.
|
| Would love to chat more about some of these ideas--shoot me
| an email!
| samstave wrote:
| Those could be plugins... or expose an api for people
| build own on data they auth to do so...
| zswaff wrote:
| Just thinking some about your note about change orders.
| Honestly super interesting. We have been working on detailed
| audit history which I think can start to help here but really
| only scratches the surface of what you need to improve.
|
| Also thinking more about
|
| > show a report on the history of the projects' intent
|
| This is a really interesting one and is actually part of where
| we started with Dart. When I was leading engineering at my last
| job I was in the habit of taking screenshots of our gantt every
| week or two so that I could manually flip through the photos
| and see how the timeline had evolved. Obviously there's no way
| that's the best solution there. We don't solve this right now
| but we plan to.
| samstave wrote:
| A change order is going to be either the change of a
| nameplate on a room, or a change to an entire layout of a
| project.
|
| Regardless of the innocuous ones being a nameplate - to as
| drastic as floorplate change... here is where I see the
| master:
|
| Project is El Camino Hospital - I was TPM.
|
| Innocuous:
|
| installing all LV net port in the MRI room, the LV consultant
| read the symbols wrong and cabled all the cables that were
| going INTO the room, on the EXTERNAL facing wall FROM the
| room.
|
| Reviewed plans were correct... change order was obv in budget
| review as we determined fault... but they had to eat that
| cost - we had to eat that TIME... as Siemens needed ports
| live on their private stuff to commission their equip on
| their end... SO even though that starts small... we can track
| all of this now against INTENT of proj budget and scope, and
| see in full how certain butterfly effects affect the result
| from the intent... (+plus data... blah blah)
|
| Bigger ones can be a relocation of a piece of equipment that
| requires extra structural... same but bigger ripple.
|
| Others can be Owner desires...
|
| But - if you could advertise a change request/order via a
| system and just have each trade reply in appropriate format
| when they are impacted - then incorporate that to the core
| project - then push back a summary with a simple "approve
| this interpretation, your responsibility and your first born
| shack, check the box"
|
| Now you can graph all the interrelated charges, parties,
| approvals - but you push out summary change reports and
| simply ensure each stake check yeah. else voice issues and
| just keep a fn journal of the process for litigation if it
| were to come to that (which is often)
|
| (OH Yeah in addition to project timeline impact --- change
| orders always include unexpected labor, equipment,
| negotiations, permits, blah blah blah -- they are CRITICAL
| money devils if plagued by them... this is what sets all
| stellar engineering firms apart... so if change mitigation is
| imbued in the process organically... $$$ - continuously align
| understanding of intent with the outcomes is wise...
| zswaff wrote:
| Loving the detailed product feedback, thanks so much. We'll
| process and take action to improve asap. Email also sounds
| great, support@ or zack@ whenever you like.
|
| Sounds like HN was rate limiting you from posting all of your
| feedback? That's awesome hahaha
|
| In short term, some quick thoughts
|
| - Right now you only have one draft, and you can access it
| through the 'Resume draft' button in the top left
|
| - We're planning to overhaul this accept/reject system for
| subtasks soon--a lot of the issues you're experiencing are very
| common, so we need to do better
|
| - Also planning to improve onboarding flow there, thanks for
| the suggestion
|
| - Not actually sure why you're being prevented from making that
| task, sorry for the trouble! We'll check it out right away.
| Tooltips are a great idea, we'll need to do more there
|
| edit: formatting
| samstave wrote:
| Cleaning garage today - so come back to respond when need a
| break. (or think of an edit)
|
| EDIT:
|
| HAHAH -- I clicked on INBOX like three fn times... the orange
| S above stole my upper perfiferal focus, so the draft button
| was invisible or subconsciously appeared greyed out and not a
| botton... - need a draft underneath INBOX, not above it - the
| same way every email side menu works.
| hoskdoug wrote:
| Was it a deliberate decision to go with a name that clashes with
| a top 20 programming language?
| jack_riminton wrote:
| Even as a dev I hadn't really thought about it clashing with a
| programming language, therefore I doubt most other people would
| care
| RamblingCTO wrote:
| That's the first thing that came to my mind though
| geniium wrote:
| First thing that came to my mind too.
| eightnoteight wrote:
| * changelog feature is pretty great
|
| * one other feature I have always wanted at my previous company
| is about transparency of how many tasks a person is doing, that
| way I don't have to bother him multiple times on whether a
| certain task is picked or not
|
| * similarly for monthly planning or weekly planning, the
| stakeholders involved in planning is often the team itself and
| humans are not that great in remembering what all promises they
| made over the last week. i wish there was some way for all
| stakeholders even outside teams to be notified of planning events
| and add items to the planning event agenda
| miladmalek wrote:
| Thanks for the ideas!
|
| - Having more information on team workloads with suggested
| opportunities for rebalancing is something we're already
| thinking about. I suppose this is a related idea to have
| general visibility on amount or total volume (with sizes
| included) of tasks a person is doing in team/public spaces.
|
| - We'll think harder about this other idea on involving outside
| stakeholders more. More opportunities for sharing lists and
| docs with external stakeholders is already something we're
| tinkering with so I can definitely see this fitting in as well.
| infocollector wrote:
| Anyone who has used this and notion - any pro/con for using this
| vs. notion for PM? (Notion these days also has an AI component)
| miladmalek wrote:
| We've used Notion as well pretty extensively and actually have
| an integration in Dart with them. From my current experiences
| Notion AI is mostly helpful for brainstorming, document
| editing, and search (for example to help with finding relevant
| info across your company docs). Their PM tool is more of an
| open-sandbox approach that doesn't always have the most robust
| features in any particular direction but is very flexible and
| easy to try out different templates for.
|
| We have a bit of a stronger emphasis on personal and team task
| management with less of a sandbox-feel and more of an
| opinionated-PM feel, and our AI features are in line with that
| with stuff like task property filling and roadmap suggestions.
| We also have similar capabilities in AI doc editing (although
| Notion and other tools are definitely more fleshed out there if
| your use case is primarily writing).
|
| That being said it's a very quickly evolving landscape and I
| wouldn't be surprised if things change in the near future.
| Would love to hear anyone else's perspective as well who've
| either tried both or used our integration.
| arnold0912 wrote:
| AI feature list seems incredible, going to check it out today!
| zswaff wrote:
| Sweet, let us know how it goes if you give it a try!
| bestlauren wrote:
| I really like the duplicate detection feature; that would save me
| so much time and prevent pain down the line with product backlog
| refinement.
| zswaff wrote:
| Yep, this is one of our most popular features. That's exactly
| our goal--catch problems like this ahead of time. We're
| actually also working on other stuff for automatic backlog
| refinement but not ready yet.
|
| If you get a chance to try, we'd love to hear how the duplicate
| detection stacks up against your expectations
| bestlauren wrote:
| I love the duplicate detection feature; this would save me so
| much time!
| e12e wrote:
| Looks interesting. Does it do progress forecasting based on
| historical task velocity, without the need for manual estimation?
|
| (See eg: https://screenful.com )
|
| Ed: Does it support importing historical data, and/or external
| data sources like GitHub, Jira, Trello etc - for current projects
| and to set a baseline grounded in data? Like historical velocity?
|
| Ed2: I guess so? From the welcome email:
|
| > Once you're logged in, you can import your tasks from most
| other apps and let us know if you need support with that
| miladmalek wrote:
| Dart will help you with forecasting based on historical task
| velocity, yes. See AI timelining features for example, and also
| with task property filling if you use sizes or date ranges.
| That being said forecasting is notoriously difficult and like
| humans, AI isn't perfect there today.
|
| There's definitely a lot more we can do here though and we have
| some plans to dive in more. I haven't looked into screenful
| before, thanks for sharing!
|
| Ed: just saw your edit, yes you can import from other tools at
| https://app.itsdart.com/?settings=import
| mrcwinn wrote:
| "We grew frustrated with X." And yet every one of these AI
| features is coming to Linear, Asana, Jira, etc. So what would
| justify the switching cost?
| miladmalek wrote:
| I guess it would be justified if the particular implementations
| we focus on (like AI roadmap planning or duplicate detection)
| are of interest to you. I'd agree that they are probably coming
| to other tools that want to adapt eventually, but by and large
| they aren't there yet! I think we take a markedly more AI-
| forward philosophy than other tools so I'm honestly not 100%
| sure that they even will fully adapt in a timely manner-
| especially the large incumbents. Most of those are still
| focused on finding different creative ways to put chatbots in
| their products instead of meshing useful AI features in a more
| streamlined way.
|
| Also admittedly this isn't a product-reason, but at least for
| now as a relatively small startup we're able to offer a lot
| more custom support and strategizing with any team that
| switches over. I can't promise we won't eventually run into the
| same bottlenecks as the large companies eventually, but for now
| I'm confident our service is better and some of our users have
| expressed that as one of the primary motivators for keeping
| with Dart.
|
| Ultimately, if you're happy with your current system and don't
| see your team becoming more efficient on Dart then it's
| probably not worth the switch. However it doesn't take long for
| any extra hours saved each week to start adding up.
| bendecoste wrote:
| This looks very cool. In every company I've worked at the biggest
| hurdle with these tools is getting the entire team to use them
| and keep them up to date. I think that gen AI could provide tons
| of value with that problem (because most teams have members that
| don't update their tickets)
|
| A while ago I had the idea that gen AI would probably be able to
| maintain a jira/kanban board of choice without people in your
| company being required to update it. You could track this based
| off what was going on with other company activity. E.g., looking
| at what people are saying in Slack, what commits a developer has
| pushed to GitHub, etc, what someone said they would do on a zoom
| call, etc.
|
| Have you guys thought about any similar features in this space?
|
| Congrats on the launch!
| zswaff wrote:
| Yep, keeping tickets updated is one of the biggest problems
| with PM today and was part of the genesis for how I got started
| working on this. We do already have an integration with GitHub
| that will automatically sync updates, as well as an integration
| with Slack that helps users turn messages into tasks. The Slack
| one doesn't really do anything automated though, just helps the
| two systems talk when you want them to. The vision to connect
| all workspaces for seamless and automatic updates (like from
| Zoom calls as you mention) is something we've thought about and
| is quite appealing.
|
| The biggest challenge with that is that every time we make an
| integration based on feedback or requests we see far less usage
| out of it than pretty much any other feature in the product. It
| turns out that people like to suggest tons of integrations but
| rarely take the time to set them up in practice. We can
| probably do a better job with that, and with giving more value
| to users sooner from the integrations.
|
| Big picture, though, your vision totally aligns with ours:
| minimize (or zero out) the time that ICs need to spend updating
| stuff while maximizing the info available for PMs etc. in the
| tool. We plan to do a better and better job with this over
| time.
| trollbridge wrote:
| Do you have any free or discounted plans for nonprofits? Monday
| offers this and one of our nonprofit partners uses it.
| miladmalek wrote:
| Yes, and we've done some great work with nonprofits before. We
| still need to make that visible somewhere on the updated
| website but shoot me an email: milad@itsdart.com
| rexreed wrote:
| How are you going to reach the market of project managers? Have
| had some experience with this in the past to mixed results.
| miladmalek wrote:
| Most of our users today have just been word of mouth, so we
| need to work harder on flexing our marketing muscles. I'd love
| to hear more about your experiences reaching PMs specifically.
|
| Early on we reached out to various PMs on Linkedin and set up
| some calls that way but that was more for feedback and
| direction-setting than anything else. I think most of our
| growth with larger teams has started from the founder or
| C-level as opposed to PMs but that might be a sign we need to
| improve more here.
| ck_one wrote:
| Can you share some insights about your dev process? App locks
| pretty slick! Congrats!
|
| What allows you to ship quickly and in high quality?
| zswaff wrote:
| Yeah, great question. This is something that I have spent a lot
| of time thinking about. Honestly probably worth a whole blog
| post/PG-style essay at some point. Shipping and iterating at
| top speed is honestly paramount for a startup.
|
| Probably the biggest thing we've done has been to focus on
| hiring for people that can and will move quickly, and building
| a company culture around that. Our team is very strong, works
| unusually hard, and ships quickly at all costs. For a startup,
| I would recommend hiring startuppy people rather than engineers
| with 20y experience at the magnificent 7.
|
| One of our most counterintuitive/provocative/potentially wrong
| practices has been to deemphasize automated tests. I'll
| probably catch a lot of flak for this but I think that spending
| a bunch of extra time writing thorough tests for something that
| will change next week when we iterate and improve is a waste of
| time. In my experience you can make up for this with thorough
| manual testing by someone (e.g. founder, product owner) that
| really, really cares about the product and UX. This is part of
| the overall "do things that don't scale" ethos and as the
| product has matured we have started to evolve to a higher level
| of test coverage. tl;dr I recommend less time spent engineering
| tests pre-PMF.
|
| Otherwise we generally follow established best practices.
| Review fast, ship fast, roll forward. Extreme ownership and
| responsibility to ICs. Use Dart for project management to save
| time ;)
|
| Overall, I think the first thing is most important: build a
| team and a culture that moves uncomfortably and unusually fast
| and breaks things (then fixes them immediately).
| lifeisstillgood wrote:
| I think project mgmt software suffers from the same problem as
| project managers - they think they should be in charge, visible
| and front and centre
|
| I disagree (esp about the human ones)
|
| I want project management that fills itself in based on work
| people do, not work people tell the tool they have done.
|
| I am dubious that putting ticket numbers in commit messages is
| needed - let alone IRC, emails, phone calls meetings, vendor
| orders placed etc
|
| I am not even sure how to do this - I think it's to do with
| defining the outcomes via automated tests (Are there five servers
| available for use by the X team, as opposed to does the blue
| button send onclick)
|
| I wish you luck but honestly saving the PM time filling in a
| report that simply "announces" things are done without being
| measuring actual results is ok but not the quantum leap. But I am
| grumpy today so perhaps I am Being too pessimistic
|
| But still - solve "achieve these automated testable goals" might
| be the right answer - but way harder as you don't control any of
| that surface area
| zswaff wrote:
| Really appreciate the feedback, doesn't come off grumpy at all.
| In fact I think it's pretty well in line with our vision.
| Beyond automated reports, we're just generally focused on
| saving users time/headache. The conflict between the people
| doing the project and the project manager is very, very common:
| the former doesn't want to spend time updating the tool, and
| the latter wants/needs centralized updates/status. In the
| future we absolutely plan to solve this better than we do now.
|
| It's super compelling to hear about different pain points like
| this. The idea of automating acceptance criteria is indeed
| tough but absolutely possible to imagine. This is outcome-
| oriented in the best possible way!
| ftkftk wrote:
| If you are using different versions of GPT, in particular GPT-4,
| how are you controlling costs once users' projects/backlogs grow
| in complexity?
| zswaff wrote:
| Great question. We're taking an unorthodox approach here.
|
| I don't know of too many tools that are offering unlimited free
| calls to OpenAI, which means that all cool LLM-enabled features
| are price-gated or premium or otherwise limited. It's a bummer
| to restrict that value. Our bet is that LLM pricing will follow
| a Moore's Law-style pattern, at least for a while, that will
| mean that we can offer better and cheaper LLM-enabled features
| over time. So in short, we're subsidizing some of the costs now
| on a longer-term bet.
|
| That said, we can be smart about how we do things technically.
| We embed, compress, and omit stuff as much as possible to
| minimize tokens.
|
| Also, we actually just completely fail to handle some things
| (something like reprioritizing a backlog of 10k tasks just
| wouldn't work for us right now) so we do hard cap some actions.
| ftkftk wrote:
| Regarding embeddings - I am assuming you are using ada-002 or
| have you moved on to 3-small already? Do you have a
| particular strategy for migrating embedding models other than
| re-embed the whole dataset? And lastly, what is/are your
| vector store(s) of choice? I am not quite sure of your scale
| but I have found that north of 50 million vectors a lot of
| the current options get a bit weak in the knees, especially
| if you index and query concurrently at high rates.
| zswaff wrote:
| - Still on ada-002, planning on migrating later this week
| actually
|
| - Current plan is to re-embed everything but I'm very open
| to better ideas there haha. Is there a better way?
|
| - I've heard some similar stuff but we haven't run into it
| yet. What are you working with?
| keeptrying wrote:
| The problem with this kind of system, is it takes away the
| interesting parts of the job for employees.
|
| I mean, who the hell wants to do a list of tasks created by an
| AI?
| rgblambda wrote:
| >repetitive PM chores like cleaning up the backlog or drafting
| changelog updates.
|
| This is something PMs are supposed to do?
| navane wrote:
| That's a lot of software for 50 minutes per week
| debarshri wrote:
| Best project management tool, IMO, is the tool that has no
| projects to manage.
|
| We tried JIRA, Linear, now we are using Github projects. It is
| too much effort to keep it up to date when things are fast
| moving. Everything becomes stale.
|
| At first look, it feels like this too will suffer the same.
|
| The best tool in this space should eliminate the need of
| specialized PM.
|
| PS. Github project has become surprising better in recent times.
___________________________________________________________________
(page generated 2024-02-20 23:00 UTC)