primitivesuave

It seems like they started with the conclusion they wanted (i.e. shareholders should bear financially responsible for the emissions of their investments) and designed a study that would confirm it. A more interesting and meaningful comparison would be of households and lifestyle/consumption habits.

The "N bots posting X times/second" isn't a very meaningful statistic. A system's reliability is mostly characterized by its performance under stress.

You're correct that it can't make things meaningfully worse. Even if he got the maximum charge for obstruction of justice (5 years), it would be an insignificant risk for someone facing 120-150 years.

I could actually see this being a viable strategy - create enough evidence and research burden for the prosecution so as to maximize the debate around admissibility of evidence and opportunities for limiting the trial's scope.

Perhaps he goes on to start the world's first decentralized cigarette currency? :)

I was referring to the general notion of someone who can satisfy exorbitant bail requirements and has two law professors for parents - legal counsel in such a case may mistakenly assume that their client is sophisticated enough to avoid obvious mistakes.

I see your point, but would more categorize his actions as a "messiah complex" belief in his own invincibility, reinforced by years of experience running the FTX charade. I doubt he was ever rationally examining the odds of pulling it off.

They're going to study SBF in law school some day. In the same way a computer science professor would emphasize automated testing to catch edge cases, a legal professor will emphasize that no matter how intelligent and well-connected your client seems, they may still be suffering from "edge case" delusional behavior that destroys their case.

This is the same guy who dialed into tech conferences and made provably false statements after the public collapse of FTX and after his parents (both law professors) had retained counsel. And yet, nobody thought to explain to him how easily he might get caught up on witness tampering and obstruction of justice charges if he continued making public statements.

His bail was secured by ~$4m in collateral. If he flees from trial, the people who bailed him out would be held legally responsible for the full $250m.

This is just a collection of static ChatGPT prompts and an output aggregator - I really don't see any novel research or real-world applicability of this aside from potentially scaffolding a project. One of the assumptions baked into the coding prompt is that you will never need to use an external library or API [1] which is an inherent requirement in most practical use cases.

> ## Code: {filename} Write code with triple quoto, based on the following list and context.

> 1. Do your best to implement THIS ONLY ONE FILE. ONLY USE EXISTING API. IF NO API, IMPLEMENT IT.

1. https://github.com/geekan/MetaGPT/blob/main/metagpt/actions/...

The most innocuous explanation is someone in management asking "if someone tells us (via API request) to send an SMS to a certain number, and we don't send that message because we're 99% sure they're being defrauded, how do we communicate that to the customer, and what about that 1% chance we're wrong?"

Although the most likely explanation is someone in management going "hold on, we have to spend money to build these safeguards, and we're going to make less money after we deploy them?"

This is actually one of the better use cases of generative AI that I've seen so far. The killer feature that would get me to purchase a $40 report is to see the competitor landscape - this is a big reason why entrepreneurs pay for Crunchbase Pro.

One annoyance I have with market research sources today is having to first select a broad industry, then narrow down the thousands of companies within that industry by quantifiable metrics (employees, fundraising, etc). It would be so much more useful to find the most semantically-relevant companies to the one I'm researching, and maybe even see an auto-generated feature comparison chart.

Creating something (writing a book, recording a song, etc) is a conversion of time (your only finite resource) into something of value (maybe only to you). It also turns out that having a profit motive and IP protection around creating valuable things is a fundamental requirement for having a creative industry to begin with. It's also what drives the free market to determine which creations are even valuable to begin with.

I remember seeing this project at an earlier stage a while back - really cool to see the progress you've made here.

I used to teach computer science to kids, and I had middle school kids setting up their own website with GitHub pages and buying their first personal domain. Firebase is also pretty amateur-friendly (although certainly more bloated than its early days) and I've seen novices deploy real-time apps with authentication, storage, server-side functions, etc.

For professional software work, the dominating reason I favor a decoupled architecture is so I can have a frontend team and a backend team. I want the frameworks and existing libraries/processes/tools to lean on, so I can minimize how much engineering time is spent on reinventing wheels.

This is a neat concept and worked great for a German conversation. Adding a pronunciation/text-to-speech element would make it a lot more useful for beginners and is especially important for any tonal languages. It might also be helpful to support spaced repetition, i.e. be able to mark an AI response as useful and then be prompted to recall it later on. There might even be some clever prompt engineering to get GPT to try favoring certain words/phrases in its response based on proficiency level.

From a business standpoint, I would suggest focusing on one language, and targeting people who are learning it with the AI + courseware. One of the GitHub founders started chatterbug.com specifically to teach German, and they delivered an exceptional experience that was easily worth $100/month (including video chats with a German coach). Now they've taken that successful model and expanded to other languages. If I could get the same conversation going with an AI, that's great, but the real secret sauce to any online learning play is the content.

Just for context to the non-climbers - climbing a v8-v9 is like being able to run a 5 minute mile or play golf with a zero handicap, and climbing at that level in your 50s is even more impressive.

This is only true if you are using dedicated IPs with SES (~$30/month each). However, SES is pretty good at managing IP reputation even for the shared IP service, and I haven't had any issues in over 10 years of building systems on top of it.

HubSpot recently launched something to do this by providing a Lambda runtime to generate components within the CRM dashboard.

https://developers.hubspot.com/docs/cms/data/serverless-func...

If you look at medical EMRs (essentially a CRM where patients are customers and salespeople are doctors) there is a similar situation as the CRM market, where a few big players dominate the scene with crappy software and feature creep (i.e. Epic = Salesforce). Based on the clean UI you have so far, you're definitely on the way to building a better user experience for CRMs, but at the end of the day you have to sell this to a salesperson, just like people selling better EMRs have to sell to hospital administrators who don't care how great the code is.

Consider an organization spending $10k/year on HubSpot. They're also spending at least a million a year on salaries/commissions for their sales staff. Optimizing software costs addresses 1% of the total spend, while making sales people more efficient optimizes the other 99%. In general, if your target customer is sales management, I would pitch support/customization contracts to streamline sales organizations when you're picking up initial customers. This would also likely bring you perspective on the wide range of customizations out there.

Congrats on launch! I completely agree with the wisdom of launching early as soon as some basic functionality works. If I had to suggest the lowest-hanging fruits for your roadmap, it would be email automation and lead enrichment.

I'm an engineering consultant for various sales tech startups which operate within the ecosystem of "build a HubSpot/Salesforce/Freshworks/etc integration for a specific type of sales organization". What I would love to see in an open-source CRM is an easy way to bring these integrations directly into the CRM - imagine a `crm.json` file which configures a CRM web app and imports custom code modules from a central repository, similar to `package.json` and NPM.

A big issue end users bring up with Salesforce/HubSpot is the high cost, especially for sales organizations which only need the core features you have in your demo today (track leads, deals, companies, etc) but have to buy a seat for each salesperson. A managed service for a hosted CRM without feature/usage/seat limitations would be an easy sell if you can reliably ingest existing CRM data and provide some level of integrations/customization.

For healthcare applications, do you sign a BAA with the customer or deploy on-premise? I could see this being extremely useful to health app developers if there is a secure way to handle patient data stored in a session replay.

I love this. I'm frequently experimenting with the "syntactic sugar" possibilities of the TypeScript type inference system, and am truly in awe of your implementation (as well as the excellent and thorough commenting).

Part of what's holding it back is expensive multi-year environmental reviews for logging projects. https://www.youtube.com/watch?v=li_Rv4t2htc

I'm an NRI/US citizen, have worked/traveled in the country extensively, and have shared my thoughts on HN many times regarding my country of origin. There is a prevalent "Indian apologist" mentality both inside and outside the country, which applauds anyone spinning the "happy people living with less" or "beautiful ancient culture" montage and shouting down anyone trying to put the preponderance of societal problems into perspective.

Pretty much every tech company is built on a stack of SaaS products (Google suite, CRM tool like HubSpot, etc), and it seems like a lot more innovation is going into integrating/standardizing their APIs.

If you're just focusing on making a pure developer tool, a good example of successful execution is cronitor.io.

The case document is especially interesting: https://www.washingtonpost.com/documents/a68b42f4-d5dc-4ff9-...

First let me say that the minimal interface you've designed for the dashboard is lovely. But also consider that as you talk to potential customers for this, the hardest sell is "send your data to a third party" and the easiest sell is "here is something that will save your developer's time". The biggest problem it seems like most people are facing now is getting their data transported between services.

Imagine finding an ancient/medieval text discussing evolution by natural selection, the indivisible nature of the atom, heliocentricity, and the existence of a unified theory of physics, all written by an author who lived in the 1st century BCE. Such was the discovery of 15th century manuscript collector Poggio Bracciolini, who discovered a surviving copy of De Rerum Natura (On the Nature of Things [1]) in a Benedictine monastery. The lost work of Lucretius inspired many Enlightenment thinkers and led people to challenge the orthodoxies of that time.

[1] https://en.wikipedia.org/wiki/De_rerum_natura

This is a very neat implementation and explanation of a disjoint set algorithm: https://en.wikipedia.org/wiki/Disjoint-set_data_structure