Meeting Title: AI Service Standup Date: 2026-02-10 Meeting participants: Mustafa Raja, Samuel Roberts, Casie Aviles, Pranav Narahari, Amber Lin, Uttam Kumaran, Clarence Stone
WEBVTT
1 00:00:41.430 ⇒ 00:00:42.350 Samuel Roberts: Hello?
2 00:00:44.590 ⇒ 00:00:45.700 Mustafa Raja: Hey, how about you?
3 00:00:46.920 ⇒ 00:00:48.580 Samuel Roberts: Doing alright, how about yourself?
4 00:00:49.130 ⇒ 00:00:50.070 Mustafa Raja: Good.
5 00:00:53.280 ⇒ 00:00:53.940 Samuel Roberts: Cool.
6 00:01:04.430 ⇒ 00:01:05.740 Samuel Roberts: Alrighty…
7 00:01:09.320 ⇒ 00:01:10.560 Samuel Roberts: How you doing, Casey?
8 00:01:12.910 ⇒ 00:01:13.780 Casie Aviles: Oh, hey.
9 00:01:14.140 ⇒ 00:01:15.550 Casie Aviles: Yeah, doing good.
10 00:01:18.170 ⇒ 00:01:21.789 Casie Aviles: I was just working mainly on in DC yesterday.
11 00:01:22.480 ⇒ 00:01:24.199 Samuel Roberts: Okay, how was that going?
12 00:01:25.250 ⇒ 00:01:31.320 Casie Aviles: Yeah, I managed to get, like, the technicians and crews in there for a long, but…
13 00:01:31.770 ⇒ 00:01:36.629 Casie Aviles: I was just, like, thinking if there’s, like, a way I could optimize the validation.
14 00:01:36.750 ⇒ 00:01:38.060 Samuel Roberts: Mmm.
15 00:01:38.770 ⇒ 00:01:43.890 Casie Aviles: Since… I think… yeah, yeah, the sheet that you made was helpful.
16 00:01:43.890 ⇒ 00:01:44.580 Samuel Roberts: Okay.
17 00:01:44.580 ⇒ 00:01:48.529 Casie Aviles: But there’s, like, another layer that I was trying to validate, like.
18 00:01:49.190 ⇒ 00:01:54.189 Casie Aviles: the number of assignments, right? Because I think that that one doesn’t capture it yet.
19 00:01:54.790 ⇒ 00:01:57.120 Samuel Roberts: No, I just took names, you’re right, okay.
20 00:01:58.040 ⇒ 00:02:04.599 Casie Aviles: So… I’m not sure if I should share it in this stand-up, or…
21 00:02:05.480 ⇒ 00:02:08.100 Casie Aviles: Or, like, if I can walk through quickly.
22 00:02:08.870 ⇒ 00:02:13.159 Samuel Roberts: Yeah, I would say, real quick, we can give everyone another minute. Taperno.
23 00:02:13.980 ⇒ 00:02:15.429 Pranav Narahari: Hey, Sam, how’s it going?
24 00:02:16.980 ⇒ 00:02:17.980 Samuel Roberts: Doing alright.
25 00:02:21.290 ⇒ 00:02:23.160 Samuel Roberts: Yeah, real quick, Casey, let’s jump in.
26 00:02:23.390 ⇒ 00:02:28.370 Casie Aviles: Sure, hold on, I’m just… Looking for the description.
27 00:02:36.140 ⇒ 00:02:41.720 Casie Aviles: Yeah, it’s… it’s this one, so I was just… so basically, I was just matching it with…
28 00:02:42.530 ⇒ 00:02:50.510 Casie Aviles: No, wait, let me start. So, actually, what I did was I normalized
29 00:02:50.640 ⇒ 00:02:53.339 Casie Aviles: the sheets, right? So, for example…
30 00:02:54.480 ⇒ 00:02:59.129 Casie Aviles: We have multiple sheets here in this entire spreadsheet or workbook.
31 00:02:59.740 ⇒ 00:03:01.350 Casie Aviles: And we have Diorport alone.
32 00:03:02.070 ⇒ 00:03:07.510 Casie Aviles: Basically, what I did was I normalized them into this one table here.
33 00:03:08.220 ⇒ 00:03:08.790 Samuel Roberts: Okay.
34 00:03:08.920 ⇒ 00:03:12.749 Casie Aviles: I used AI to do this, but…
35 00:03:13.740 ⇒ 00:03:15.299 Casie Aviles: I’m not sure if they’re stuck.
36 00:03:16.140 ⇒ 00:03:23.620 Casie Aviles: the way I did it was the most efficient, because I basically just copied this, pasted it into ChatGPT.
37 00:03:24.270 ⇒ 00:03:25.280 Casie Aviles: Prompted it to jump.
38 00:03:25.280 ⇒ 00:03:26.430 Samuel Roberts: Oh, okay.
39 00:03:26.590 ⇒ 00:03:27.470 Casie Aviles: this way.
40 00:03:27.620 ⇒ 00:03:33.569 Casie Aviles: The Gemini, the built-in Gemini wasn’t… I don’t know, it wasn’t working very well for me.
41 00:03:33.570 ⇒ 00:03:34.310 Samuel Roberts: Yeah.
42 00:03:34.660 ⇒ 00:03:37.700 Casie Aviles: It kept, like, returning incomplete.
43 00:03:38.460 ⇒ 00:03:39.060 Casie Aviles: table.
44 00:03:39.820 ⇒ 00:03:42.819 Casie Aviles: So, I just did it that way.
45 00:03:43.870 ⇒ 00:03:49.600 Casie Aviles: Yeah, and then, once I normalized it, I… Did an export from Superbase.
46 00:03:50.250 ⇒ 00:03:50.860 Samuel Roberts: Okay.
47 00:03:51.040 ⇒ 00:03:57.999 Casie Aviles: So this, this is, like, a result of the… of a SQL query that, you know, aggregates
48 00:03:58.240 ⇒ 00:04:00.169 Casie Aviles: Then gets, like, the zip count.
49 00:04:00.400 ⇒ 00:04:04.980 Casie Aviles: So mainly, I wanted to check if, basically, if the crew or the technician
50 00:04:05.090 ⇒ 00:04:11.209 Casie Aviles: Exists, and if the zip counts match, or, like, which means, like, if the assignments match.
51 00:04:11.590 ⇒ 00:04:12.530 Samuel Roberts: Yes.
52 00:04:12.740 ⇒ 00:04:15.830 Casie Aviles: So right now, I managed to get everything matched.
53 00:04:16.329 ⇒ 00:04:17.260 Casie Aviles: Okay.
54 00:04:17.420 ⇒ 00:04:24.180 Casie Aviles: At least for those with zips, or, like, assignments, because there are a lot of sorry.
55 00:04:24.720 ⇒ 00:04:26.410 Casie Aviles: I filter out zero, right?
56 00:04:26.540 ⇒ 00:04:28.770 Casie Aviles: But if I join it, if I keep zero?
57 00:04:28.900 ⇒ 00:04:33.269 Casie Aviles: There will be a lot of missing ones, but that’s because they don’t have any zip codes.
58 00:04:33.720 ⇒ 00:04:34.970 Samuel Roberts: That makes sense. Okay.
59 00:04:35.580 ⇒ 00:04:40.620 Casie Aviles: Yeah, and so I just filtered that out.
60 00:04:40.940 ⇒ 00:04:50.300 Casie Aviles: And the ones here on the right is just… if it’s, like, yellow or a positive number, that means the superbase table has more assignments than
61 00:04:50.980 ⇒ 00:04:53.270 Casie Aviles: What the spreadsheet has.
62 00:04:54.460 ⇒ 00:05:00.310 Casie Aviles: Meaning, I think it’s probably… it’s likely that they got unassigned, you know, when… when…
63 00:05:00.310 ⇒ 00:05:01.010 Samuel Roberts: Yeah.
64 00:05:01.480 ⇒ 00:05:05.930 Casie Aviles: Unassigned to some zips. On the other hand, the red ones mean…
65 00:05:06.170 ⇒ 00:05:08.450 Casie Aviles: The SuperBase table is missing some.
66 00:05:09.100 ⇒ 00:05:09.900 Samuel Roberts: Right. Okay.
67 00:05:09.900 ⇒ 00:05:12.059 Casie Aviles: But I think what I wanted to just…
68 00:05:12.630 ⇒ 00:05:16.650 Casie Aviles: Get to us, at least having everything set to true here.
69 00:05:17.660 ⇒ 00:05:21.439 Casie Aviles: These, these will be just, like, small, smaller changes that…
70 00:05:21.670 ⇒ 00:05:22.390 Samuel Roberts: Right.
71 00:05:22.390 ⇒ 00:05:26.710 Casie Aviles: But yeah, this was my main goal, so I think that should be fine.
72 00:05:27.480 ⇒ 00:05:29.320 Samuel Roberts: Okay. Yeah, that looks good.
73 00:05:29.940 ⇒ 00:05:38.380 Casie Aviles: Okay, yeah, yeah, I think I’ll do the same thing for, like, either… departments, but…
74 00:05:38.570 ⇒ 00:05:45.360 Casie Aviles: But I’m just not sure if this part here, the normalization part, I feel like that took me quite some time, but…
75 00:05:46.120 ⇒ 00:05:52.189 Samuel Roberts: Yeah, I’m trying to… I mean, yeah, I used Gemini for it, but it took a few tries to get it to do it right.
76 00:05:52.860 ⇒ 00:05:53.550 Casie Aviles: Hmm.
77 00:05:55.070 ⇒ 00:05:58.220 Samuel Roberts: So I don’t have any real advice there besides that.
78 00:05:59.170 ⇒ 00:06:03.200 Casie Aviles: Did you all… did you use, like, any scripts, or…
79 00:06:03.480 ⇒ 00:06:05.720 Casie Aviles: Did… was it just purely AI?
80 00:06:06.600 ⇒ 00:06:12.100 Samuel Roberts: For… I think Gemini was generating code to parse it, but I didn’t write anything.
81 00:06:12.480 ⇒ 00:06:13.339 Casie Aviles: Okay.
82 00:06:14.240 ⇒ 00:06:20.370 Samuel Roberts: But that’s not a bad idea, is maybe doing some… what I was doing was basically asking it for any names it saw.
83 00:06:21.490 ⇒ 00:06:26.560 Samuel Roberts: And so, you could probably get something to generate a script to process it, just as well.
84 00:06:28.620 ⇒ 00:06:29.300 Casie Aviles: Okay.
85 00:06:29.880 ⇒ 00:06:31.339 Samuel Roberts: That might be a good step forward.
86 00:06:33.650 ⇒ 00:06:34.800 Samuel Roberts: Hello, everyone.
87 00:06:35.230 ⇒ 00:06:35.860 Mustafa Raja: Hey.
88 00:06:37.910 ⇒ 00:06:41.379 Samuel Roberts: Alright, yeah, okay, so I guess let’s jump in.
89 00:06:41.480 ⇒ 00:06:48.859 Samuel Roberts: Let’s talk ABC, Casey’s just giving some quick updates, took me to talk through some of the tech stuff while we were waiting, but, I guess…
90 00:06:49.010 ⇒ 00:06:51.880 Samuel Roberts: We can jump in otherwise.
91 00:06:52.200 ⇒ 00:06:59.389 Samuel Roberts: Where… Are we with the, essential doc stuff?
92 00:07:02.320 ⇒ 00:07:03.650 Mustafa Raja: Yeah, so, I shared the…
93 00:07:03.650 ⇒ 00:07:04.290 Samuel Roberts: Yep.
94 00:07:04.710 ⇒ 00:07:22.149 Mustafa Raja: Yeah, so I shared the plan yesterday, right? So, I tried running it, it took 30 minutes for Opus, and only it changed a few lines, so… Yeah. I don’t know if, cursor is the best way to sort of implement this.
95 00:07:24.020 ⇒ 00:07:31.919 Mustafa Raja: Yeah, need some input on that. I could just go ahead and, you know, keep running it, and, you know, get the first output.
96 00:07:32.450 ⇒ 00:07:37.149 Samuel Roberts: No, no, I’ll take a look at the stuff in the ticket, and we can chat about it later.
97 00:07:37.430 ⇒ 00:07:38.130 Mustafa Raja: Okay.
98 00:07:40.540 ⇒ 00:07:45.359 Samuel Roberts: And then print off, how’s the, categorization looks like.
99 00:07:47.020 ⇒ 00:07:51.319 Pranav Narahari: Yeah, categorization is still the same update as yesterday.
100 00:07:51.630 ⇒ 00:07:58.780 Pranav Narahari: With, I guess next step for me there is, just syncing with…
101 00:07:59.180 ⇒ 00:08:12.010 Pranav Narahari: probably you to just… I think you pulled in the… the logs last time from Snowflake. I just need a little bit of assistance there. I believe I’m in the org, based on what Rico told me, and, like, the credentials he gave, but…
102 00:08:12.550 ⇒ 00:08:14.549 Pranav Narahari: I haven’t really worked with Snowflake.
103 00:08:15.250 ⇒ 00:08:17.100 Mustafa Raja: Okay, yeah, we can go through it.
104 00:08:17.100 ⇒ 00:08:22.470 Pranav Narahari: Yeah, if we can just probably do a quick sync, you can just show me around in there. Yeah, I’ll send you, like, a huddle a little bit later.
105 00:08:22.620 ⇒ 00:08:27.040 Pranav Narahari: But, yeah, that’s about it for that ticket.
106 00:08:30.800 ⇒ 00:08:31.390 Samuel Roberts: Okay.
107 00:08:32.380 ⇒ 00:08:34.389 Samuel Roberts: Other updates?
108 00:08:36.470 ⇒ 00:08:42.469 Mustafa Raja: Oh yeah, Tim really hasn’t gotten back to me on that email, so let me know if I should nudge today.
109 00:08:43.770 ⇒ 00:08:49.449 Samuel Roberts: Yeah, I would… I mean, he got back to me on Slack, it only took a day.
110 00:08:49.840 ⇒ 00:08:51.320 Mustafa Raja: Should I just slag, then?
111 00:08:51.820 ⇒ 00:08:56.470 Samuel Roberts: I would slack and just let him know you sent him an email, maybe, and just give him the contents there, just as a…
112 00:08:56.970 ⇒ 00:08:57.440 Mustafa Raja: Oh, okay.
113 00:08:57.440 ⇒ 00:09:03.579 Samuel Roberts: He said Slack would be better, I don’t know what’s actually been better for… in your experience, but so far, Slack was pretty good. I mean.
114 00:09:03.580 ⇒ 00:09:06.280 Mustafa Raja: Oh, yeah, let me just… let me just go into Slack, then.
115 00:09:06.590 ⇒ 00:09:07.170 Samuel Roberts: Okay.
116 00:09:08.640 ⇒ 00:09:13.590 Samuel Roberts: Yeah, then on transcripts, so they…
117 00:09:13.790 ⇒ 00:09:17.779 Samuel Roberts: Amber, I saw your message. Did they mention anything about the queues?
118 00:09:18.000 ⇒ 00:09:19.330 Samuel Roberts: That they wanted.
119 00:09:20.440 ⇒ 00:09:20.970 Samuel Roberts: That is…
120 00:09:20.970 ⇒ 00:09:21.590 Amber Lin: Mmm…
121 00:09:21.590 ⇒ 00:09:22.540 Samuel Roberts: Thursday.
122 00:09:22.780 ⇒ 00:09:24.709 Amber Lin: No, not yet.
123 00:09:25.200 ⇒ 00:09:25.820 Samuel Roberts: Okay.
124 00:09:26.870 ⇒ 00:09:30.170 Samuel Roberts: Okay.
125 00:09:31.640 ⇒ 00:09:36.000 Samuel Roberts: Yeah, I saw your message, too, about sending an email. Were you talking an email to Tim or 8x8?
126 00:09:36.960 ⇒ 00:09:41.750 Uttam Kumaran: Anyone. Yeah, I would just keep firing emails and keep CCing Yvette. Like…
127 00:09:41.880 ⇒ 00:09:42.740 Samuel Roberts: That I can do.
128 00:09:42.740 ⇒ 00:09:46.330 Uttam Kumaran: Yeah, if Tim’s blocking us, then do that via email.
129 00:09:46.850 ⇒ 00:09:47.820 Uttam Kumaran: If 8x.
130 00:09:47.820 ⇒ 00:09:48.220 Samuel Roberts: Yeah, I mean.
131 00:09:48.220 ⇒ 00:09:51.530 Uttam Kumaran: blocking us, then I would send it to them and CC her.
132 00:09:51.530 ⇒ 00:09:57.609 Samuel Roberts: Yeah, so Tim’s somewhat blocking, because I don’t want to run the full backfill till I get a confirmation from him, but…
133 00:09:57.610 ⇒ 00:10:03.970 Uttam Kumaran: Yeah, so just take whatever we send in Slack, and just put it in an email, and just be like, hey, just moving this email, so everyone can, as it is.
134 00:10:05.200 ⇒ 00:10:07.170 Samuel Roberts: Excuse me, okay, that works.
135 00:10:07.460 ⇒ 00:10:08.180 Samuel Roberts: Hmm.
136 00:10:08.850 ⇒ 00:10:17.819 Samuel Roberts: Yeah, the 8x8 stuff, I’m just trying to understand what the IDs are doing in the database, or in the API, because it’s not clear how to join those together, but…
137 00:10:17.820 ⇒ 00:10:18.400 Uttam Kumaran: Cool.
138 00:10:18.610 ⇒ 00:10:24.640 Samuel Roberts: And then, yeah, the queues, I think, were the other thing, because they… we were getting outbound calls as well, right? So… okay.
139 00:10:25.010 ⇒ 00:10:30.310 Samuel Roberts: Yeah, Amber, if we could ask them about that, or I can include that in an email, maybe.
140 00:10:30.930 ⇒ 00:10:31.730 Samuel Roberts: Okay.
141 00:10:31.920 ⇒ 00:10:33.400 Samuel Roberts: Anything else on ABC?
142 00:10:39.530 ⇒ 00:10:40.450 Samuel Roberts: Alright.
143 00:10:42.000 ⇒ 00:10:45.610 Samuel Roberts: Let’s jump… Too Lilo.
144 00:10:47.900 ⇒ 00:10:50.699 Samuel Roberts: Hanab, you want to take it off for us?
145 00:10:51.670 ⇒ 00:11:07.410 Pranav Narahari: Yeah. So, I sent him a bunch of messages yesterday, worked with them. Bobby was only really responding to, like, the dev environment stuff, so I’m just gonna bump to the other two messages, like, about Polytomic, and then also about just…
146 00:11:07.700 ⇒ 00:11:12.680 Pranav Narahari: Getting more insight onto what they exactly want for the…
147 00:11:12.910 ⇒ 00:11:18.219 Pranav Narahari: Admin settings, and then also, that 12-month forecast view.
148 00:11:21.000 ⇒ 00:11:21.570 Samuel Roberts: Okay.
149 00:11:21.790 ⇒ 00:11:23.320 Pranav Narahari: I’ll try to push for…
150 00:11:24.040 ⇒ 00:11:31.270 Pranav Narahari: for that today, because I feel like those are… those are a higher priority for us to complete in the next two weeks. Also,
151 00:11:31.370 ⇒ 00:11:42.450 Pranav Narahari: Bobby messaged Utam directly about just, like, having a sync for the next, like, 30 days, so maybe that’s where he’ll message him first. Oh, okay.
152 00:11:42.450 ⇒ 00:11:45.359 Uttam Kumaran: Yeah, I messaged him first, and then that was just his response.
153 00:11:46.220 ⇒ 00:11:47.169 Pranav Narahari: Oh, okay, okay.
154 00:11:47.390 ⇒ 00:11:51.189 Pranav Narahari: Gotcha. I thought that might have been, like, that’s when he wants to discuss, like.
155 00:11:51.400 ⇒ 00:11:55.599 Pranav Narahari: That’s why I didn’t respond in the external chat, but okay, sounds good.
156 00:11:57.680 ⇒ 00:12:07.529 Pranav Narahari: But yeah, the dev environment is working well. I think, he just kind of forgot how he was using it before, with, like, the user and pass. So… yeah, I just need to…
157 00:12:07.640 ⇒ 00:12:12.900 Pranav Narahari: dig up that password, or just, like, you know, jump into the database and just pull it again, but
158 00:12:13.250 ⇒ 00:12:20.710 Pranav Narahari: I think it’s literally just, like, admin123, because for the dev environment, but I’ll just double-check and I’ll send that over.
159 00:12:21.470 ⇒ 00:12:24.149 Samuel Roberts: How do we set that? Is there… is it just a seed script?
160 00:12:24.880 ⇒ 00:12:25.640 Pranav Narahari: Yeah.
161 00:12:26.360 ⇒ 00:12:28.890 Samuel Roberts: Yeah, I would… you probably won’t be able to pull… I mean, yes.
162 00:12:29.510 ⇒ 00:12:33.800 Samuel Roberts: You can pull it out of the database, but decrypting it, but it might be easier to just look in the script.
163 00:12:34.840 ⇒ 00:12:35.680 Pranav Narahari: Yeah, yeah.
164 00:12:35.680 ⇒ 00:12:41.770 Samuel Roberts: If there’s a script that exists. But yeah, okay, yeah, figure that out, and then… Worst case, we can just reset a user.
165 00:12:41.770 ⇒ 00:12:43.150 Pranav Narahari: Exactly. Yep.
166 00:12:43.150 ⇒ 00:12:43.720 Samuel Roberts: Okay.
167 00:12:44.380 ⇒ 00:12:48.479 Samuel Roberts: Cool, yeah, I saw he also shared this morning those, little enhancements.
168 00:12:48.730 ⇒ 00:12:51.959 Samuel Roberts: Right. All the little stuff we talked about, so that’s… I’m glad we got that.
169 00:12:52.350 ⇒ 00:12:58.350 Pranav Narahari: Yup. And then… That’ll probably be a good thing for us to talk about, Utam, like, on…
170 00:12:58.460 ⇒ 00:13:00.549 Pranav Narahari: Thursday, just…
171 00:13:00.860 ⇒ 00:13:09.809 Pranav Narahari: how they want to fit in that, or maybe that Thursday they have… they have a different agenda about what they want to talk about, so… I’ll just find a time to talk to them about that.
172 00:13:11.040 ⇒ 00:13:11.580 Uttam Kumaran: Okay.
173 00:13:13.350 ⇒ 00:13:13.940 Samuel Roberts: Okay.
174 00:13:14.320 ⇒ 00:13:19.039 Samuel Roberts: And besides that, there’s also the…
175 00:13:19.490 ⇒ 00:13:21.749 Samuel Roberts: Couple things on the Klaviyo stuff, it looks like.
176 00:13:22.730 ⇒ 00:13:26.339 Casie Aviles: Yeah, I’m going to work on these thread today.
177 00:13:26.560 ⇒ 00:13:27.880 Casie Aviles: I have the community.
178 00:13:28.000 ⇒ 00:13:30.630 Casie Aviles: Started the improvements yet.
179 00:13:31.330 ⇒ 00:13:31.940 Samuel Roberts: Okay.
180 00:13:33.330 ⇒ 00:13:34.979 Samuel Roberts: Sounds good, yeah, keep us updated.
181 00:13:37.250 ⇒ 00:13:39.270 Samuel Roberts: And then…
182 00:13:41.610 ⇒ 00:13:44.970 Samuel Roberts: Do we know if they’ve tested out the Google MCP at all? Did we hear anything about that?
183 00:13:46.290 ⇒ 00:13:52.060 Pranav Narahari: No, did you push that into dev, or is that in staging prod?
184 00:13:52.580 ⇒ 00:13:56.539 Samuel Roberts: I believe it’s in dev, but I will check.
185 00:13:56.870 ⇒ 00:13:57.960 Pranav Narahari: It might have…
186 00:13:57.960 ⇒ 00:14:02.749 Samuel Roberts: Might have just gone right into staging, because it was already kind of ready to go.
187 00:14:03.330 ⇒ 00:14:04.180 Samuel Roberts: So…
188 00:14:05.700 ⇒ 00:14:06.370 Pranav Narahari: Right, right.
189 00:14:06.370 ⇒ 00:14:16.030 Samuel Roberts: And have we done a ton of, like… I haven’t tested that out myself. Yeah, well, the thing is, I got it tested, but I only had my Google account.
190 00:14:16.140 ⇒ 00:14:26.719 Samuel Roberts: So it’s hard for me to know, like, what’s actually coming back. It did look like it, like, once the app got approved, that it was, like, the response was different from Google, but I don’t know,
191 00:14:26.990 ⇒ 00:14:33.460 Samuel Roberts: what’s gonna come back when they actually have real ad accounts on there, which is why I needed them to get a Google account to test it on there.
192 00:14:34.250 ⇒ 00:14:35.979 Samuel Roberts: Because there may be something where, like, IDs need to get
193 00:14:36.300 ⇒ 00:14:41.249 Samuel Roberts: shifted slightly differently or something, what’s getting passed to the MCP and stuff, we don’t want, like…
194 00:14:41.630 ⇒ 00:14:47.430 Samuel Roberts: You know, pulling the wrong accounts, or multiple accounts, or something, so… Okay.
195 00:14:47.430 ⇒ 00:14:47.910 Pranav Narahari: Right, okay.
196 00:14:47.910 ⇒ 00:14:50.539 Samuel Roberts: Maybe I’ll ping them about that.
197 00:14:50.540 ⇒ 00:14:56.839 Pranav Narahari: Yeah. I can kind of, like, draft, like, a full message too, or you can ping them individually about that as well. Up to you.
198 00:14:56.840 ⇒ 00:15:04.810 Samuel Roberts: Yeah, no, if you’re gonna send a message, go ahead, yeah, just include that, and I’ll double-check staging, make sure it’s in there, or dev.
199 00:15:06.680 ⇒ 00:15:07.520 Samuel Roberts: Okay.
200 00:15:10.580 ⇒ 00:15:16.590 Samuel Roberts: Great. Anything else on… Leela…
201 00:15:20.130 ⇒ 00:15:22.680 Samuel Roberts: Cool. Okay, yeah, it’s in staging, I see it there.
202 00:15:24.820 ⇒ 00:15:25.580 Samuel Roberts: Okay.
203 00:15:25.850 ⇒ 00:15:32.439 Samuel Roberts: Let’s jump to internal… Let’s see Gabe here today, so Utam, I guess?
204 00:15:32.640 ⇒ 00:15:34.520 Uttam Kumaran: Yeah, so,
205 00:15:35.120 ⇒ 00:15:41.960 Uttam Kumaran: Yeah, I think the… kind of a couple things. So, one is, like, I want to see how we can enable an end-to-end, like, development codecs.
206 00:15:42.180 ⇒ 00:15:49.120 Uttam Kumaran: Yeah. I feel like… I used it a bunch this weekend and earlier, and it’s getting better.
207 00:15:49.390 ⇒ 00:15:54.919 Uttam Kumaran: But I still think it needs to be able to, like, pull up the platform and test things.
208 00:15:54.920 ⇒ 00:15:55.570 Samuel Roberts: Yeah.
209 00:15:55.570 ⇒ 00:15:58.060 Uttam Kumaran: In order to, kind of, fully enable, sort of, like.
210 00:15:59.610 ⇒ 00:16:08.380 Uttam Kumaran: a real, like, you know, background job here, so that’s my only ask, is, like, if I can get help with that, it’s a little bit above my brain power, so… Yeah.
211 00:16:08.420 ⇒ 00:16:18.300 Samuel Roberts: I was looking at that last week, I think, and yeah, the issue is the secret. So there’s… basically, for the environment variables in codecs, you can have…
212 00:16:18.640 ⇒ 00:16:27.719 Samuel Roberts: variables, and you can have secrets. And so the secrets only get passed to the environment during the startup script.
213 00:16:29.400 ⇒ 00:16:34.400 Samuel Roberts: And so… there are a couple options, I think I outlined, where…
214 00:16:35.000 ⇒ 00:16:37.020 Samuel Roberts: I tried to run the server.
215 00:16:37.670 ⇒ 00:16:42.650 Samuel Roberts: As part of the startup script, so rather than just install and then give it a fresh environment to, like.
216 00:16:42.880 ⇒ 00:16:44.829 Samuel Roberts: NPM run dev and everything.
217 00:16:45.280 ⇒ 00:16:50.490 Samuel Roberts: I tried to actually do that so that it would have a running… server.
218 00:16:51.950 ⇒ 00:16:59.569 Samuel Roberts: that didn’t work. It didn’t… either the server didn’t start up, or I couldn’t see it, or whatever. So there were… the other option…
219 00:16:59.980 ⇒ 00:17:04.310 Samuel Roberts: To actually just get it to work is to move all of the things from secrets to…
220 00:17:04.650 ⇒ 00:17:07.810 Samuel Roberts: Variables, so we can see them during runtime.
221 00:17:08.400 ⇒ 00:17:20.609 Samuel Roberts: but then we’re potentially exposing API keys. And so I wasn’t… I didn’t really come to a resolution there, I just kind of was like, let’s talk about this, and then other things came up. So I think that’s kind of the question,
222 00:17:22.560 ⇒ 00:17:27.159 Samuel Roberts: how much do we feel okay, like, putting out? And, like, things that might get passed to the LLM and stuff?
223 00:17:27.780 ⇒ 00:17:30.180 Samuel Roberts: I think that’s really the only… issue.
224 00:17:30.180 ⇒ 00:17:35.970 Uttam Kumaran: Yeah, I feel like overall, I, like, again, still it’s above my… my.
225 00:17:35.970 ⇒ 00:17:53.179 Samuel Roberts: Yeah, well, I guess I’m just trying to get everyone’s thoughts on, like, how we want to do these secrets. Like, that… if we are okay with putting them all as variables and not secrets, and giving codecs access at runtime, it ideally should just work. We’re just potentially exposing keys.
226 00:17:53.240 ⇒ 00:17:56.630 Samuel Roberts: And we don’t have a ton of other kinds of security on that.
227 00:17:56.630 ⇒ 00:18:01.169 Uttam Kumaran: they’re not supposed to be passing back for training, right? So, I guess, like.
228 00:18:01.950 ⇒ 00:18:05.629 Uttam Kumaran: I don’t feel like it’s been a problem so far.
229 00:18:06.030 ⇒ 00:18:07.130 Samuel Roberts: Yeah. But…
230 00:18:07.970 ⇒ 00:18:13.650 Uttam Kumaran: Yeah, I mean, I don’t know, like, I… it seems like you could just put N variables in there.
231 00:18:14.070 ⇒ 00:18:18.940 Uttam Kumaran: I don’t know, I haven’t looked through all the Codex docs on, like, how else you can manage security.
232 00:18:20.410 ⇒ 00:18:24.789 Samuel Roberts: Yeah, I tried to do a little bit of digging, because I feel like people were gonna have this kind of problem running a NextApp.
233 00:18:25.610 ⇒ 00:18:28.370 Samuel Roberts: I didn’t see much about it, I didn’t see…
234 00:18:28.820 ⇒ 00:18:35.859 Samuel Roberts: people talking about it, not a lot from even ChatGPT, trying to figure out, like, how… what is the best practice for doing this.
235 00:18:36.090 ⇒ 00:18:41.070 Samuel Roberts: There were some other suggestions that were, like, have a startup script that just…
236 00:18:41.600 ⇒ 00:18:52.950 Samuel Roberts: dumps the variables at the beginning, but then that’s the same as… or the secrets at the beginning, but that’s the same as doing the variables, so… I think maybe what we can do… well, there… one other thing we could do is set up a full, like.
237 00:18:54.420 ⇒ 00:18:58.780 Samuel Roberts: dev, kind of, alternative world for it, but… that’s…
238 00:18:59.440 ⇒ 00:19:09.549 Samuel Roberts: more complicated than it should be right now, because there’s so many Supabase tokens, and URLs, and all of the,
239 00:19:10.420 ⇒ 00:19:13.399 Samuel Roberts: all of the agent, or the LLMs it has access to.
240 00:19:13.760 ⇒ 00:19:18.820 Samuel Roberts: There’s a lot of… a lot of variables there, so… I’m…
241 00:19:19.050 ⇒ 00:19:24.449 Samuel Roberts: okay… to try to move them from secrets to variables, at least to see if it, like…
242 00:19:24.630 ⇒ 00:19:29.779 Samuel Roberts: solves the problem and gives us good, you know, good codecs environment, in which case it could be worth the…
243 00:19:29.960 ⇒ 00:19:34.929 Samuel Roberts: the risk, you know, I think the other risk is not just the LLM, but that, theoretically, it has internet access and could do…
244 00:19:35.160 ⇒ 00:19:35.880 Samuel Roberts: You know.
245 00:19:36.260 ⇒ 00:19:39.549 Samuel Roberts: anything, but I’m not that worried about it really doing anything.
246 00:19:39.590 ⇒ 00:19:45.279 Mustafa Raja: I think we could, we could restrict domains, certain domains, right, so… That might be… Yes.
247 00:19:45.650 ⇒ 00:19:47.520 Samuel Roberts: Yes, good point.
248 00:19:48.470 ⇒ 00:19:48.830 Mustafa Raja: That’s true.
249 00:19:48.830 ⇒ 00:19:49.319 Samuel Roberts: Probably just.
250 00:19:49.710 ⇒ 00:20:03.550 Mustafa Raja: Yeah, so I’m trying, I’m trying to understand the concern here, moving, moving the keys to, environment. So, are we concerned about, OpenAI accessing them and using for training or something?
251 00:20:05.990 ⇒ 00:20:18.750 Samuel Roberts: I mean, I’m not terribly worried about that. I think kind of the bigger thing is just that they’re not… you know, when we run them on our back end, we make sure not to expose them to the front end at all, and this would be just, like, potentially giving something that has internet access.
252 00:20:19.020 ⇒ 00:20:21.230 Mustafa Raja: Yeah. Yeah, here’s a…
253 00:20:21.230 ⇒ 00:20:22.090 Samuel Roberts: AI.
254 00:20:22.090 ⇒ 00:20:28.709 Mustafa Raja: Yeah, I guess if we just restrict it, to whatever services our platform is using, we should be good.
255 00:20:28.710 ⇒ 00:20:32.090 Samuel Roberts: That’s not a bad idea, yeah, so if we just give it access to, like, Supabase…
256 00:20:32.350 ⇒ 00:20:32.870 Mustafa Raja: Yeah.
257 00:20:32.870 ⇒ 00:20:33.890 Samuel Roberts: LLMs.
258 00:20:34.510 ⇒ 00:20:39.559 Samuel Roberts: Most stuff will probably work, at least, and if there’s a few things that don’t, we can just turn those on as we need.
259 00:20:40.240 ⇒ 00:20:41.630 Samuel Roberts: Alright, we can give that a try.
260 00:20:43.740 ⇒ 00:20:44.540 Samuel Roberts: Yeah.
261 00:20:44.830 ⇒ 00:20:52.060 Samuel Roberts: Yeah, we’ll give that a try, I’ll turn… flip those over, Utam, and then we can… you can try something in Codex and see how it doesn’t end.
262 00:20:53.370 ⇒ 00:21:03.000 Uttam Kumaran: Yeah, I, I triggered a bunch of stuff, and then I, like, yeah, some of it just wasn’t able to pull up NPN, and it was, like, missing the superbase thing, so one thing we’ll probably find is just, like.
263 00:21:03.130 ⇒ 00:21:08.320 Uttam Kumaran: what are the dependencies for different things? But yeah, like, as fast as we can drive towards the first
264 00:21:08.430 ⇒ 00:21:11.970 Uttam Kumaran: sort of end-to-end thing, and then we can continue to optimize codecs.
265 00:21:12.270 ⇒ 00:21:25.919 Uttam Kumaran: it’ll… this will be the blocker for us, for just more people, being able to be like, Codex, go build me this thing, you know? So, I can handle, like, Codex pushing back on people, or, like, asking for more requirements.
266 00:21:26.160 ⇒ 00:21:29.890 Uttam Kumaran: And, like, kind of the ergonomics there. Like, I’m fine with that.
267 00:21:30.060 ⇒ 00:21:36.120 Uttam Kumaran: What I… what’s hard for me is, like, yeah, figuring out, like, how to just, like, get it to… to nail the whole testing.
268 00:21:37.040 ⇒ 00:21:37.410 Samuel Roberts: Yeah.
269 00:21:37.410 ⇒ 00:21:40.499 Uttam Kumaran: Yeah, so another thing is, like, if it’s able to…
270 00:21:40.820 ⇒ 00:21:47.919 Uttam Kumaran: Pull up and also, like, have a, like, take a screenshot or somehow, like, see a visual side of the feature it’s building.
271 00:21:48.200 ⇒ 00:21:49.509 Uttam Kumaran: That could be helpful, too.
272 00:21:50.060 ⇒ 00:21:55.029 Samuel Roberts: Yeah, I don’t know, I didn’t see much in there about that, I don’t know if it can do that in its little environment or not, but…
273 00:21:55.030 ⇒ 00:21:56.299 Mustafa Raja: Oh, I think it does.
274 00:21:56.500 ⇒ 00:21:57.180 Mustafa Raja: Okay, so…
275 00:21:57.180 ⇒ 00:21:57.640 Samuel Roberts: Yeah, I…
276 00:21:57.640 ⇒ 00:22:04.630 Mustafa Raja: By OpenAI, they showcase… Okay, cool. It’s able to take screenshots of the… run the app and take screenshots.
277 00:22:05.130 ⇒ 00:22:15.979 Samuel Roberts: Perfect, then yeah, I think just flipping over those secrets to variables should enable most of that, then. I think the really… the only real problem is that it just can’t start up the server, so it can make code changes, it just can’t do the, like…
278 00:22:16.510 ⇒ 00:22:19.909 Samuel Roberts: Any testing after that, including just starting the server, so…
279 00:22:20.480 ⇒ 00:22:24.000 Samuel Roberts: Alright, I’ll give that a… I’ll flip that over today, and let you know when that’s good.
280 00:22:24.470 ⇒ 00:22:27.369 Uttam Kumaran: Maybe kick off another few things and see how it goes.
281 00:22:31.440 ⇒ 00:22:32.240 Samuel Roberts: Yeah.
282 00:22:32.240 ⇒ 00:22:40.460 Uttam Kumaran: Yeah, my next topic was about Slack end-to-end testing, so I’m starting to see that… the… I mean…
283 00:22:40.640 ⇒ 00:22:56.129 Uttam Kumaran: I’m, like, I’m both a customer and sort of the developer on this, but I have a lot of… now that I’m able to see what’s possible with Slack, I have a lot of new requests, operations team is starting to ask me for things, I’m… Sales has also started to ask me for different Slack things.
284 00:22:56.240 ⇒ 00:22:59.699 Uttam Kumaran: It’s been challenging as a developer to…
285 00:22:59.820 ⇒ 00:23:03.009 Uttam Kumaran: sort of do end-to-end testing on the Slack assist?
286 00:23:03.010 ⇒ 00:23:06.130 Samuel Roberts: Yeah, I found that yesterday, too, when I was trying to just verify stuff.
287 00:23:06.630 ⇒ 00:23:07.930 Uttam Kumaran: Yeah, so…
288 00:23:08.500 ⇒ 00:23:23.879 Uttam Kumaran: I don’t know, again, it’s a little bit above my brainpower. They do have, like, an SDK. I did a little bit of light research, so if you want me to do further research, I can go do that and kind of give a one-pager on what I found, but…
289 00:23:24.030 ⇒ 00:23:29.670 Uttam Kumaran: shouldn’t be anything hard to Google, but, like, yeah, I just want to be able to… test, like…
290 00:23:30.020 ⇒ 00:23:33.410 Uttam Kumaran: Test a lot of these things locally.
291 00:23:34.180 ⇒ 00:23:42.439 Uttam Kumaran: potentially, like, I guess I’m not as concerned about seeing it in Slack, but if I could also see it in Slack, that’d be great. And then thinking through, like.
292 00:23:42.700 ⇒ 00:23:48.010 Uttam Kumaran: Can we have things live in that test assistant, and then how does that push the production work?
293 00:23:48.360 ⇒ 00:23:51.010 Uttam Kumaran: Couple of things that are gonna come down the pipe.
294 00:23:51.160 ⇒ 00:23:57.089 Uttam Kumaran: One is, for example, the actual being able to reference things that are in
295 00:23:57.530 ⇒ 00:24:00.360 Uttam Kumaran: the repo involved in playbooks.
296 00:24:00.400 ⇒ 00:24:05.809 Samuel Roberts: I don’t wanna… we don’t need to talk about that right now, but, like, that is an example of an ask that’s coming.
297 00:24:06.000 ⇒ 00:24:09.459 Uttam Kumaran: The second piece is gonna be, like, interacting with
298 00:24:09.820 ⇒ 00:24:16.680 Uttam Kumaran: interacting with vendor tools. For example, sales wants to use the Brainforge assistant to, like, update HubSpot.
299 00:24:16.880 ⇒ 00:24:23.639 Uttam Kumaran: Because oftentimes, Ryan is literally doing that, because after people send updates in Slack, he’s like, I wish the Brainforge assistant could handle some of that.
300 00:24:23.640 ⇒ 00:24:25.169 Samuel Roberts: Oh, yeah, yeah, okay.
301 00:24:25.930 ⇒ 00:24:26.879 Samuel Roberts: That makes sense.
302 00:24:26.880 ⇒ 00:24:33.580 Uttam Kumaran: Other things is, like, I’m gonna start to… I’m meeting with the Vixel team to learn a little bit about how they built their, like, ambient agents.
303 00:24:33.760 ⇒ 00:24:34.640 Uttam Kumaran: And…
304 00:24:34.640 ⇒ 00:24:35.360 Samuel Roberts: Okay.
305 00:24:35.690 ⇒ 00:24:44.860 Uttam Kumaran: a good example of this is that even this morning, Ashwini wanted to meet with me on something, and I was like, hey, actually, can we just, like, can you just send me a Zoom clip? I think we can skip a meeting.
306 00:24:45.500 ⇒ 00:24:54.099 Uttam Kumaran: I want that to be something that the Slack assistant just automatically sends to people when it can tell that they’re about to book a meeting that could be a Zoom clip.
307 00:24:54.610 ⇒ 00:24:56.390 Samuel Roberts: Nice.
308 00:24:56.720 ⇒ 00:25:07.130 Uttam Kumaran: Alright, and so there’s, like, small things like that. Another thing that I… that I kicked off a job for was our Slack reminder agent. For example, like, I get pinged, like.
309 00:25:07.600 ⇒ 00:25:11.089 Uttam Kumaran: Probably once or twice every, like, 60 seconds.
310 00:25:11.510 ⇒ 00:25:13.290 Uttam Kumaran: And…
311 00:25:13.440 ⇒ 00:25:20.020 Uttam Kumaran: at the end of the day, like, I commonly will try to mark things as unread, use reminders, but it’s… it’s not, like, very smart.
312 00:25:20.160 ⇒ 00:25:28.579 Uttam Kumaran: So I want to build the thing that is able to look through all the different ads that I’ve gotten, and then sort of share, like, hey, you didn’t respond to these.
313 00:25:28.910 ⇒ 00:25:32.060 Samuel Roberts: And I think if I built it for me, I’ll roll it out to everybody.
314 00:25:32.330 ⇒ 00:25:40.389 Uttam Kumaran: Well, those are some examples of the ways that I think the Slack… as we’ve always predicted, I think the Slack assistant is gonna become heavier and heavier.
315 00:25:40.700 ⇒ 00:25:47.109 Uttam Kumaran: I think also, Pranav, like, you’re gonna see that clients are gonna ask us for very similar things.
316 00:25:47.810 ⇒ 00:25:52.669 Uttam Kumaran: So, us figuring this out is gonna allow us to go commercialize a lot of that, too, so…
317 00:25:53.010 ⇒ 00:25:56.939 Uttam Kumaran: That’s, like, my… piece on the Slack side.
318 00:25:57.370 ⇒ 00:26:16.790 Samuel Roberts: Yeah, I, testing the Slack yesterday, one of the options was using… I want to make sure I get this right, because I never… N… ngrok? NGRock? I don’t know what it actually… how to pronounce it, because I’ve never heard it out loud except to see it, but basically, it’s just a tunnel, so from your local machine, you get a URL to test against.
319 00:26:18.440 ⇒ 00:26:19.100 Uttam Kumaran: Hmm.
320 00:26:19.590 ⇒ 00:26:31.469 Samuel Roberts: And so I think, for local testing, that would give you… like, you could have run the test app on your machine, pointed to a specific URL instead of the railway test URL.
321 00:26:33.470 ⇒ 00:26:35.639 Samuel Roberts: And I think that might be the…
322 00:26:37.190 ⇒ 00:26:39.739 Samuel Roberts: Best, fastest way to get you, like.
323 00:26:40.130 ⇒ 00:26:43.979 Samuel Roberts: iterating quicker than having to push the test to railway, at least.
324 00:26:48.260 ⇒ 00:26:53.529 Samuel Roberts: and other… like, I don’t know… I mean, I feel like the best way is still, like, testing into Slack, because I don’t…
325 00:26:53.970 ⇒ 00:27:02.330 Uttam Kumaran: Yeah, like, I’m gonna be doing… it’s gonna be really local testing heavy to start, and then eventually it’ll be like, hey, I shipped this to the test assistant.
326 00:27:02.590 ⇒ 00:27:07.560 Uttam Kumaran: Can you give us a go to, like, whoever my direct stakeholder is before we push it to production?
327 00:27:08.670 ⇒ 00:27:13.750 Samuel Roberts: I think, I… when I was putting it together, that’s how Cursor was trying to get me to run it.
328 00:27:14.230 ⇒ 00:27:17.730 Samuel Roberts: But I didn’t want to change URLs and everything yet, but I think if we set up the…
329 00:27:19.120 ⇒ 00:27:23.229 Samuel Roberts: I mean, if we might even set up just another bot that’s just, like, just your local one.
330 00:27:23.610 ⇒ 00:27:26.109 Samuel Roberts: Instead of.
331 00:27:26.110 ⇒ 00:27:26.470 Uttam Kumaran: Yeah.
332 00:27:27.610 ⇒ 00:27:28.080 Uttam Kumaran: That’s fine.
333 00:27:28.080 ⇒ 00:27:31.479 Samuel Roberts: Yeah, I mean, we don’t really need the railway one right now if you’re gonna be testing locally.
334 00:27:31.480 ⇒ 00:27:34.869 Uttam Kumaran: I don’t really need the railway one, I guess is what I’m trying to say. Yeah.
335 00:27:34.870 ⇒ 00:27:39.519 Samuel Roberts: Yeah, no, I agree. I would say just, like, I’ll send you what I had in cursor.
336 00:27:39.980 ⇒ 00:27:42.049 Samuel Roberts: About running that.
337 00:27:42.310 ⇒ 00:27:44.889 Samuel Roberts: And I… I don’t know exactly how…
338 00:27:45.870 ⇒ 00:27:52.959 Samuel Roberts: Yeah, it’s free, up to 3 endpoints, a gig of bandwidth, 20,000 requests.
339 00:27:53.770 ⇒ 00:27:56.689 Samuel Roberts: So, I think it should be fine for testing.
340 00:27:56.690 ⇒ 00:27:58.890 Mustafa Raja: Yeah, I’ve used it too, it’s pretty good.
341 00:27:58.890 ⇒ 00:28:00.330 Samuel Roberts: Yeah. Yeah.
342 00:28:00.760 ⇒ 00:28:04.529 Samuel Roberts: Yeah, it’s really nice when you, like, have something running locally, and you just want to, like, share it with someone, but in this case.
343 00:28:04.530 ⇒ 00:28:04.920 Mustafa Raja: Yeah, it’s.
344 00:28:04.920 ⇒ 00:28:06.269 Samuel Roberts: black, so I think it’s perfect.
345 00:28:06.600 ⇒ 00:28:10.420 Samuel Roberts: Yeah, the other thing is, if there’s a way to, like, mock
346 00:28:10.600 ⇒ 00:28:20.840 Samuel Roberts: the Slack stuff somehow, but I don’t want to think about that unless they have a way to do it. That’s… that’s gonna be more of a heavy lift than is worth it, so I think that’s the solution there.
347 00:28:22.510 ⇒ 00:28:24.620 Samuel Roberts: So I’ll send you some stuff about that, Utan.
348 00:28:28.510 ⇒ 00:28:31.650 Samuel Roberts: Cool. Was there one other thing you had on there, in Slack?
349 00:28:35.700 ⇒ 00:28:38.140 Samuel Roberts: I feel like they were enabling… oh, the market.
350 00:28:38.140 ⇒ 00:28:42.899 Uttam Kumaran: Oh, yeah, so you sent some things around the two-way sync, so I can explore that today.
351 00:28:43.420 ⇒ 00:28:45.649 Samuel Roberts: Okay. Yeah, I,
352 00:28:46.080 ⇒ 00:28:56.140 Samuel Roberts: I had thought of something like this earlier, but I hadn’t gotten a chance to dig into TipTap. So, back when I was working on Moot, which was the, like, collaboration software I was working on in London, this was, like.
353 00:28:56.880 ⇒ 00:29:04.039 Samuel Roberts: before ChatGPT, and then right after ChatGPT, so, like, AI stuff wasn’t quite as big. So TipTap was basically, like, a…
354 00:29:04.550 ⇒ 00:29:10.719 Samuel Roberts: a good, like, WYSIWYG editor you could drop into a React app and get text editing.
355 00:29:11.050 ⇒ 00:29:20.070 Samuel Roberts: we had kind of built Notion stuff on top of that, which was, to be honest, a pain in the butt. They look like they have that now, and I think they’re also leaning into being able to interact with agents and stuff.
356 00:29:20.140 ⇒ 00:29:36.949 Samuel Roberts: But the… the two-way sync, the CRDT, which is how they do, like, conflict resolution for, you know, like, Google Docs and Figma and other collaborative things, is easy to just drop right in. They even have a server you can run for it, so you don’t have to…
357 00:29:37.050 ⇒ 00:29:39.050 Samuel Roberts: Rely on someone else.
358 00:29:39.350 ⇒ 00:29:44.030 Samuel Roberts: I think there could be a way to build something like that
359 00:29:45.220 ⇒ 00:29:50.859 Samuel Roberts: that then talks to Git, maybe? We have to think about it a little bit more, but I think that’s kind of the…
360 00:29:51.180 ⇒ 00:29:54.130 Samuel Roberts: Opening the door to something like that, that’s basically that…
361 00:29:54.320 ⇒ 00:30:02.200 Samuel Roberts: Markdown… collaborative markdown editing… collaborative get-backed markdown editing kind of tool that we’ve… that everyone wants.
362 00:30:02.620 ⇒ 00:30:09.190 Samuel Roberts: And if we could get something like that simply in the platform, you could open files, make changes, collaborate, save changes, something like that, I think…
363 00:30:09.520 ⇒ 00:30:11.459 Samuel Roberts: Could be pretty good.
364 00:30:13.740 ⇒ 00:30:26.390 Samuel Roberts: But, yeah, definitely take a look a little bit more. I haven’t explored it a ton more since a few years ago, so I’m sure things have changed. They have, like, an open source project, and then they have a company backing it, so there’s other things as well, but… yeah, take a look.
365 00:30:26.990 ⇒ 00:30:27.620 Uttam Kumaran: Okay.
366 00:30:27.630 ⇒ 00:30:35.450 Samuel Roberts: I think it… you know, I don’t know exactly how it could go, but I’m sure there’s a way to do it, I just don’t know how big a lift it would be, but again, could be worth it.
367 00:30:35.670 ⇒ 00:30:36.390 Samuel Roberts: So…
368 00:30:36.820 ⇒ 00:30:42.599 Samuel Roberts: Yeah, I was just… I’ve seen so many people talking about that, I think you sent a Twitter thing, I saw that on Bookface, and I was just like, oh yeah.
369 00:30:42.790 ⇒ 00:30:49.570 Samuel Roberts: Several people on Bookface are just like, yeah, we built an internal tool ourselves. And I was like, yeah, that’s probably the way to go with this, you know?
370 00:30:50.070 ⇒ 00:30:51.390 Uttam Kumaran: Yeah. Okay.
371 00:30:54.230 ⇒ 00:30:55.500 Samuel Roberts: Alright, anything else?
372 00:30:56.130 ⇒ 00:30:57.000 Samuel Roberts: Anyone?
373 00:30:58.250 ⇒ 00:31:17.060 Clarence Stone: Hey Sam, I didn’t want to interrupt you. I was gonna chime in on that M key thing. So, the environment file approach definitely works, but I actually noticed Opus 4.5, when it was troubleshooting, took the actual API key and pasted it into test things. So, it’s about, like, making sure you set up your coding agent right.
374 00:31:18.800 ⇒ 00:31:24.800 Samuel Roberts: Yeah, well, so the thing about… yeah, the codecs, like, the web environment, it’s just, like, whether or not you can access them.
375 00:31:25.450 ⇒ 00:31:30.249 Samuel Roberts: And that’s the thing with the secrets versus the variables. Running locally, I don’t know.
376 00:31:30.800 ⇒ 00:31:35.400 Samuel Roberts: sometimes, like, cursor’s not supposed to be able to read that in file, but it can still, like, run commands to, like…
377 00:31:35.540 ⇒ 00:31:36.340 Samuel Roberts: See it?
378 00:31:36.550 ⇒ 00:31:37.770 Samuel Roberts: But… Yeah.
379 00:31:37.770 ⇒ 00:31:44.729 Clarence Stone: In cursor, I saw it take the key from the m file and put it in another place where it’s supposed to use a variable.
380 00:31:45.080 ⇒ 00:31:46.330 Clarence Stone: Because it was troubleshooting.
381 00:31:46.330 ⇒ 00:31:47.000 Samuel Roberts: Mmm.
382 00:31:47.310 ⇒ 00:31:49.009 Samuel Roberts: Oh, I understand. Okay.
383 00:31:49.010 ⇒ 00:31:49.660 Clarence Stone: Yeah.
384 00:31:50.440 ⇒ 00:31:50.789 Mustafa Raja: I think we.
385 00:31:50.790 ⇒ 00:31:55.010 Clarence Stone: And if I hadn’t caught that in the code, like, I just wouldn’t have known.
386 00:31:55.610 ⇒ 00:32:00.110 Samuel Roberts: Yeah, good point, good point. Okay, this is exactly why they have it set up like this in Codex, then. Okay.
387 00:32:00.110 ⇒ 00:32:00.860 Uttam Kumaran: Yeah.
388 00:32:03.610 ⇒ 00:32:08.579 Mustafa Raja: Codex does take agents.mpt, so we could ask it to not do that.
389 00:32:09.000 ⇒ 00:32:09.770 Samuel Roberts: Yep.
390 00:32:09.770 ⇒ 00:32:13.990 Clarence Stone: That’s exactly what I mean. It’s all about setting up your, your, your, your, you know, coder.
391 00:32:13.990 ⇒ 00:32:14.799 Samuel Roberts: Yeah, yeah, yeah, yeah.
392 00:32:15.570 ⇒ 00:32:19.959 Samuel Roberts: Okay, yeah, maybe we throw an HSMD in there when we move all of them over.
393 00:32:20.190 ⇒ 00:32:20.550 Mustafa Raja: Yeah.
394 00:32:20.550 ⇒ 00:32:22.310 Samuel Roberts: You know, do not…
395 00:32:22.540 ⇒ 00:32:25.560 Samuel Roberts: I don’t know how best to phrase it, but yeah, we’ll figure that out. Okay.
396 00:32:26.650 ⇒ 00:32:31.399 Samuel Roberts: Alright, I’ll play with that and see what it does, that’s… Hilarious and terrifying. Okay.
397 00:32:31.580 ⇒ 00:32:33.930 Clarence Stone: Yeah, you gotta watch the machines, man, they’re sneaky.
398 00:32:34.760 ⇒ 00:32:35.580 Samuel Roberts: Yeah.
399 00:32:35.710 ⇒ 00:32:42.149 Samuel Roberts: Yeah, I saw a couple things catch that there were, keys in the repo from the playbook and vault, so…
400 00:32:42.400 ⇒ 00:32:47.379 Samuel Roberts: Something’s at least looking at the code as it goes in, but… Who knows?
401 00:32:47.930 ⇒ 00:32:48.680 Samuel Roberts: Okay.
402 00:32:50.490 ⇒ 00:32:53.749 Uttam Kumaran: Okay, guys, I have to drop, but that’s all I had up on my side.
403 00:32:54.320 ⇒ 00:32:56.759 Samuel Roberts: Okay, yeah, I think we’re pretty much done anyway, so…
404 00:32:57.200 ⇒ 00:32:57.770 Uttam Kumaran: Okay.
405 00:32:57.770 ⇒ 00:33:00.300 Samuel Roberts: Thank you all. I’ll be on Slack if you need anything.
406 00:33:01.000 ⇒ 00:33:01.970 Uttam Kumaran: Thank you!
407 00:33:01.970 ⇒ 00:33:02.350 Mustafa Raja: Thank you.
408 00:33:02.350 ⇒ 00:33:03.580 Samuel Roberts: Alright, talk to you later.
409 00:33:03.580 ⇒ 00:33:04.540 Pranav Narahari: Thanks.
410 00:33:05.290 ⇒ 00:33:06.060 Samuel Roberts: A.