@NacksonD@Latesha_Byrd I would assume demand is actually very high in-between Atlanta and Charlotte. Charlotte is the second largest banking hub in the country, so the amount of people connecting via Atlanta is probably the majority of the traffic.
Deepseek v4 is an inferior product to OpenAI's platform in almost every way
Can you use Deepseek for audio to text transcriptions?
No
Can you Deepseek to create text embeddings for vector search? No
Is Deepseek as good as writing software? Also no
I would personally pay the premium for GPT-4-mini if my platform dealt with knowledge work in anyway, amazing model!
What I think most people are missing, even @TheStalwart, is we are seeing the software engineering classโs usual dogmatic tendencies play out in public.
Historically, users had no idea what the backend architecture was or where the code was hosted. However, if you've been on a software team, you know engineers historically have almost cared more about these decisions than end user experience. Then there was a lot of people who would also include political motivation in these decisions.
A few years ago, you saw a huge push for OSS people to take projects off of Github and host them somewhere else. Now people are saying its because Github is not a reliable platform, but my memory tells me the main issue with hosting your code with someone who also works with the federal government and ICE.
I think we are seeing something similar now
Choosing Claude over OpenAI/Codex maps almost perfectly with previous moral fights in software engineering.
Notice how every time you see some banner headline about token inefficiency it always in relation to Claude. Opus 4.7 objectively cost more and does worse than Codex 5.5 for engineering task. If that is becoming more and more true, why aren't we seeing SWE all jump to Codex immediately? Aren't these people supposed to be the best at weighing trade offs.
If you didn't have this dogmatic personality, it felt like were apart of a minority of SWE, maybe it's a silent majority idk.
I'm opening Dispatch to private beta today ๐
For agents, a wallet can be more than payment. It can be identity.
Dispatch gives that identity an inbox: receive email, verify accounts, catch receipts, reply, and hand off when needed.
DM me what workflow you'd run
The next problem Dispatch will solve is that of agent reputation.
I believe in the future websites will care about what agents are coming to interact with them, how can I trust this agent wont immediately be a bad actor?
Preventing bad actors on Dispatch is a top priority
Agents are becoming economic actors.
They reason, use tools, run workflows, and pay.
But their operational identity is split across wallets, inboxes, memory, and human handoffs.
Iโm building something to collapse that into a single identity layer.
Private beta opens tomorrow.
@ajlamesa It's a surprise if you had a superficial understanding of Georgia Politics
I've been saying she's going to win the primary out right for a few weeks
Supabase would've been almost perfect if they just gave you a vanilla Postgres database as the default.
The "Supabase Way" of building apps is genuinely so stupid. Before LLMs, using Supabase did help accelerate app development, I just don't see the benefit anymore.