Hacker News new | past | comments | ask | show | jobs | submit login
DoppelBot: Replace Your CEO with an LLM (modal.com)
232 points by gk1 13 days ago | hide | past | favorite | 117 comments





It's fun to think about, but if you want to really reign in the C-suite, you need to start setting up systems within companies that allow for the rank-and-file to replace bad CEOs.

We've spent the last 80 years (ostensibly) talking about how great representative government is, then we let people own a majority of voting shares so that they can run companies like dictatorships. I've worked at a company that could have remained more competitive (and done a better job of maintaining software that is critical to human life) if there had been more uptake of ideas from workers instead of simply trying to make shareholders happy.


This is called a cooperative, and it is completely legal. However, they tend not to be very successful in the marketplace, outside of a few outliers.

If you separate revenue and control (employees get to vote on decisions, but investors get the revenue), then employees will be incentivized to maintain the status quo instead making decisions to make the company successful. The employees have almost no incentive to make difficult decisions like cutting unsuccessful lines of business etc. I would be very uneasy about investing anyone's retirement savings into such an organization.

If you keep revenue and control combined (a simple share structure, where each share gets a piece of the revenue and a vote, and employees get shares when they join), then you have other paradoxes. To make this work, you need to prohibit employees from selling their shares to investors, otherwise you're back to a conventional corporation. You also need to take the shares away from employees when they leave the company, otherwise former employees become a de facto investor class that you wanted to avoid. As an employee, I would prefer to work at a company where my equity does not have these restrictions placed on it.

You may say "but if we all just tried a little harder to believe in a better world it could be possible". But it is possible now. Cooperatives work easily under existing corporate law, and are even incentivized in some places. They just aren't very successful for the reasons above. The only widespread form of cooperative is doctor's and lawyer's practices, which are legally required to be organized as cooperatives through a limited liability partnership structure.

To make cooperatives work in the broader marketplace, you'd have to force employees to accept restrictions on their equity or force savers to put their money into cooperatives managed by employees who are not incentivized to invest it well.


I've been part of a few (Canadian) cooperatives. The incentive structure is completely different, it's akin to admitting you prefer great working conditions than having a chance to chance it rich. If a company is startup-sized, anything else is a sucker's bet.

If I didn't like my working conditions, I would look for a different job.

>I would look for a different job.

I know some experienced, socially adept, smart, etc. friends who have been doing this exact thing for about 8 months now.


How is that not the same in a coop? If they don't like the conditions they'd still have to leave the job and find a new one.

In a coop everybody's job is to always be looking for a job (for the coop)?

Not sure what you're talking about, I was replying to a refutation of the following quote by the grand parent commenter:

> If I didn't like my working conditions, I would look for a different job.


because all your friends are 10s

What's your point? That my friends aren't the literal best in the world in their field of study? Of course they aren't. If that was a job requirement, there would be like 20 job openings world-wide.

What an odd comment.


Sometimes the bad jobs are crowding out the good. Like gig work crowding out traditional full-time positions with benefits, as companies prefer cheaper, on-demand labor. Slave labor needs to be outlawed at the society level, not at the individual level "why dont you just find another job".

gig work is not slave labour - slaves had no choice.

gig work is popular because the people doing it are finding that it is the best of all alternatives available to them.


"slaves had no choice"... Um, do you think most people can choose not to work?

Gig work is "popular" because wealth-driven systemic pressure is replacing FTE with gig jobs in order to fatten the bottom line.


> do you think most people can choose not to work?

they have the choice - because they have agency. Slavery is literally taking that agency away. Having agency doesn't mean being immune to the consequences.

The consequences of a slave "choosing" to not work is punishment and death.

The consequences of a gig worker choosing not to work is to claim basic benefits (but hardly will die). It's uncomfortable, and i will bet most people will choose to work, but it doesn't mean someone is forcing them.


Look at the rest of his comments. Can't reason with him, so just move on.

This "take it or leave it" problem seems to be exactly what a cooperative seeks to solve.

Of course, but being in such a choice position is already being very privileged if we take things at their global level.

What was the incentive structure, and how was your experience overall (esp. comparing to traditional companies if you’ve worked at those)?

> If you keep revenue and control combined (a simple share structure, where each share gets a piece of the revenue and a vote, and employees get shares when they join), then you have other paradoxes. To make this work, you need to prohibit employees from selling their shares to investors, otherwise you're back to a conventional corporation. You also need to take the shares away from employees when they leave the company, otherwise former employees become a de facto investor class that you wanted to avoid

This describes most “partnerships” to a varying degree. Lots of professional firms operate that way, e.g. law firms, accounting firms, engineering firms, consulting


Yes, a large percentage of of the consulting / services business that you would know the name of are organized like this. For services / consulting it makes a lot of sense.

All of your points are premised on the idea that the only purpose of a corporation is to make as much money for its shareholders as possible. But that isn't true, especially in the case of a cooperative.

Most companies make decisions based on a list of priorities: they care about their shareholders/investors first, upper management second (upper management is usually bribed by the investors with equity, aligning their interests with the shareholders, so this is maybe a distinction without a difference), rank-and-file employees third, customers fourth, and anyone not involved with the company last.

The point of a cooperative is to flip this incentive structure on its head: the customers or workers are owners, so they get prioritized. The purpose of the company isn't to dominate the markets and become the biggest company ever, it's to make a better life for the people working for and patronizing the company, instead of just the people who own it.

> As an employee, I would prefer to work at a company where my equity does not have these restrictions placed on it.

Well obviously, if you're talking about equity in a company as something you can own and profit from without ongoing labor, you are definitionally a capitalist and you will be opposed to cooperatives.

> I would be very uneasy about investing anyone's retirement savings into such an organization.

The primary purpose of a cooperative is not to make profits for passive investors. This is like judging a fish on its ability to climb a tree.


> The purpose of the company isn't to dominate the markets and become the biggest company ever, it's to make a better life for the people working for and patronizing the company, instead of just the people who own it.

That's perfectly fine, and also the reason why cooperatives are a pretty niche thing: they don't prioritize growth.


This is only half the answer. Cooperatives are niche in the US because they don't prioritize growth, and they exist in a legal and financial framework that does. If conventional corporations didn't receive massive amounts of investor cash and preferential government treatment to stay unprofitable for years or decades while scaling, one imagines coops would be more competitive.

They are investors for a reason, they expect a return. If a coop does not produce good returns, especially as compared to standard corporations, why would one expect investors to invest?

Exactly. Investors will always constitute a force against cooperatives; countries with successful coops (Spain, Italy, the Nordics, even Canada) provide other funding mechanisms to prevent private investment from dictating the entire market. The United States chooses not to do this. It is a choice, not the natural result of having a finance system.

> provide other funding mechanisms to prevent private investment from dictating the entire market.

if this funding comes from the gov't, then you're just forcing taxpayers to subsidize a less profitable funding mechanism. It's merely a wealth transfer (from taxpayers to these coop owners).


Since 1980 the top 1%’s share of national income nearly doubled, while the bottom 50%’s share fell from 20% to 13%. Real wages for most workers stagnated despite productivity rising 62% (1979-2019), ALL of that went to investors.

People who claim funding co-ops is a "wealth transfer" ignore existing policies which already funnel taxpayer money upward. Redirecting a fraction of subsidies like bailouts and fossil fuel breaks and the endless war machine... instead to worker owned models that reinvest profits locally and tie wealth to labor... that isn’t a new transfer but a correction. Co-ops reduce reliance on exploitative private equity (e.g., hedge funds buying 40% of U.S. rental homes) and ensure economic gains stay with the people generating value.

I have the controversial opinion that people who work should get money, not the people who don't work.


> people who work should get money

and they do.

> not the people who don't work

and they don't. With the exception of some basic welfare payments, which i am not opposed to, what you mentioned is what currently happens.

But you're unilaterally classifying capital as not being work would be wrong. Capital is solidified work from previous/earlier times, but unconsumed and saved.


I think they're referring to the class of people who profit from capital rather than their labour. That class profits from the labour of others. They don't have to exploit their bodies, experience, and skills (labour) to realize profits. They're the ones who, "don't work."

They're not talking about people who "don't work," and collect welfare.


Capitalists don’t work. That’s what they meant.

The US government has a lot of funding mechanisms, the economy does not rely solely on private investment. It is just that the scale of venture investment dwarfs a lot of even governmental investment in most other countries, money made from previous ventures so of course they'd deploy that capital.

They don't prioritize growth for the same reason healthy cells don't prioritize growth.

You need to be cancer or a public corporation to think growth is the most important thing.


There isn't any value judgement here. There are not more cooperatives because they don't prioritize growth. Who are you angry at?

This oft-repeated analogy to biological systems doesn't make much sense, human individuals biologically stay homeostatic with nature, but human groups have grown exponentially, building civilization as you know it. If we didn't prioritize growth, we'd still be hunter gatherers.

>However, they tend not to be very successful in the marketplace, outside of a few outliers.

What would you consider to be successful, and what is marketplace naming here?

Cooperatives are going to be the main thriving organizational structure when the political agenda in place favors them, and the very same statement apply for resource omni-capture oligopolies. Look at who have hands on the political schedule, and that’s all nailed.

Living in France, I never worked for any structure that would provide any equity (not saying this doesn’t exist at all). But France still has a very different non-wage labour costs compared to US (or common-law in general I guess), so employment tradeoffs are very different.


There are a few food companies that are partially or completely employee owned that I'd consider "very successful". But this business model is not typical and examples are rare.

https://inequality.org/article/5-workerowned-food-firms-riva...


Why do folks think that is? Are there some US-specifics that make them less likely?

For example, the Mondragon Corporation, https://en.wikipedia.org/wiki/Mondragon_Corporation, in the Basque region of Spain is pretty huge (70,000 employees) and I watched a short news documentary about Mondragon and in general the employees there sounded legions happier than corporate drones in other companies.


And most of those companies are low three-digit employee counts. I mean, I like King Arthur and tend to visit their store when I'm in the area, but it's only a relatively significant employer by the standards of Norwich, VT.

> then employees will be incentivized to maintain the status quo instead making decisions to make the company successful.

Therein lies the problem.

Capitalists believe these two things are exclusionary to one another. They believe that you cannot have an unsuccessful product line, or "status quo" and still be successful. They want to see comfy upward trending lines on their charts and graphs.

But nobody stopped to consider that;

1) If the business is operating with positive cashflow, and,

2) If the business is meeting it's commitment to consumers, shareholders, and employees, and,

3) If the business has no direct existential threat to it's market share, then;

That business is successful. By your own definition. Stop trying to take over the world. It's selfish, unsustainable, and pedantic. Support your workers. Well. Put value back into the economy. Stop being a greedy POS.


There is no secret conspiracy stopping cooperatives from working. They simply do not supply a large portion of the products and services that people want, due to the fact that they are by definition not set up to do so.

Who are you angry at? You are yelling into the wind.


If all these 3 are true, then that seems to me like a definition of a monopoly.

Specifically the condition 3 is practically never true in well functioning market. If you're successful, you'll be copied and you'll have competitors, so you always have to be a step ahead. Maintaining the status quo is how you slowly become irrelevant


Some people just like going to their favorite chain restaurant, bar, corner store, etc.

Not everything is a fight.


It seems pretty bizarre to pick restaurants and bars as an example, that's insanely competitive market. It's quite literally a fight for survival, restaurants go out of business all the time here. (Central Europe).

I currently watch the exact case from the comment above in our street - new coffee place showed up, and the old one, which kept the status quo for very long time is now left empty as people go to the new one instead.

It's also a good example of how growing and merging into larger chains is a working defense strategy for the market share threat.


> they tend not to be very successful in the marketplace

...because other companies boycott coops.


Strange that economic power is still handed down the generations like political was in the days of kings

+1. I do think there's certain efficiency gains in concentrating decision-making in companies, but nepotism and hereditary power are strange characteristics to keep from feudalism

A company is an authoritarian regime, not a democracy. Real power is kept within a small circle. Since employees can ascend, it might look like a meritocracy, but that is only allowed if it strengthens the regime.

> We've spent the last 80 years (ostensibly) talking about how great representative government is, then we let people own a majority of voting shares so that they can run companies like dictatorships.

I think it’s basically admitting that the right dictator, if they are there based on merit, can really perform better than someone who wins by their skill at campaigning. Forget voting shares - even when a single person doesn’t have that amount of control, you have a small board elected by shareholders who is essentially electing a dictator.

Personally I think most companies would do far worse if rank and file employees selected the CEO. But I haven’t thought much about it. Curious what ideas HN has on this aspect.


The advantages of democracy are a peaceful transition of power and non-violent outlets for political action. That is to say, you don’t have to worry about a succession war when the king dies, and if people don’t like what the government is doing they tend to advocate and vote rather than try to overthrow it.

These advantages don’t really apply to companies. Your life isn’t tied to a company the way it is to your government, so if you don’t like how they chose the leader or what they’re doing, you can leave.


> Your life isn’t tied to a company

Health insurance isn't tied to a company? Your ability to survive and feed your family isn't tied to your company? Lack of opportunities in the job market? Lack of equal salary/benefits at alternative companies? Non-competes, H1-B visas....

> so if you don’t like how they chose the leader or what they’re doing, you can leave.

Just one small thing Ben: leave to work for who? Who isn't hiring CEOs and running businesses the same way as everyone else?


Please don’t quote half a sentence and then argue with it when the second half is an important qualifier.

Many of these are uniquely American problems, and the many of the people and especially the highest in power in the United States have chosen and continue to choose to bind all of that to the company you work for (or don't).

Yes, it's ironic that a democracy has exactly created the situation they lament about, the lack of universal healthcare. Unfortunately there is no political will to do it, while ironically an autocrat could cut through the red tape and get it done, as it had been enacted in several Asian countries in the last half century.

I don't see it much of an admission of anything, personally. The admission would have had to come as an answer to a question that was never asked. To me, it's more of a way to keep the ruling class in place. It used to be done through government, now it's done through corporations.

And honestly, the corporation is a far more attractive model if you're in that racket. Rioters don't burn down corporate HQs, they burn down police stations. Public sector pay sucks. The corruption would be far too obvious... well, it would have been until relatively recently.


> I think it’s basically admitting that the right dictator, if they are there based on merit, can really perform better than someone who wins by their skill at campaigning.

Sure. But also there are big differences between companies and countries/government.

Companies compete in a market. There isn't much of a marketplace for country/government. This is a check on individual power, and forces the company to perform somewhat in alignment with what we want as a society. Concentration of power (monopoly) is still an issue in a market and so we regulate the market.

The "dictator" of a company answers to both the market and to regulators, making it less dangerous as a concentration of power than an actual dictator.

Related, it is totally okay for companies to fail, and they regularly do. We are a lot less okay with governments failing.

Companies can also serve a niche. Governments need to serve their entire population. Governments are tasked with all sorts of collective action problems for diverse groups with varying opinions; companies get to pick their market. It's unclear that CEOs tell us much about a dictator trying to serve a population with a range of opinions.


> The "dictator" of a company answers to both the market and to regulators, making it less dangerous as a concentration of power than an actual dictator.

Well, kind of. On occasion they participate in regulatory capture making them really only accountable as they care to be given the circumstances.

If you have a market cornered on AI-powered, autonomous, robotic leg-equipped chainsaws, and the regulatory environment is weak because you've lobbied for the right rules and the right enforcers of those rules, you could very possibly just have a bad quarter if those chainsaws' AI sees people as trees and starts chasing them around. Maybe not even that.

Meta would be a good real-world example. It's just less interesting to imagine.


To prove you right, you can read up on the incredible giga-brained countrywide experiments by Kardelj in Socialist Yugoslavia [0]. The result being a country where no-one wanted to work, and everyone had a great standard of living (while the IMF didn't call in its loans). And then the entire country collapsed all at once under the accumulated mismanagement.

[0] - https://en.wikipedia.org/wiki/Workers%27_self-management#Yug...


It's wild to me that despite tremendous resources and 100+ years of time capitalism still kills millions of people a year with starvation and preventable diseases. But every right winger has a pet wikipedia page about a failed communist state with no critical examination of why they failed beyond "communism bad".

To clarify my stance I'm an anarchist and that page has a lot of good examples of successful worker owned collectives.


There are good critically examined rebuttals if you actually look for them beyond Wikipedia which is not designed for that purpose. I read a book recently called Socialism: the failed idea that never dies, and while it has a clickbait title, the arguments are pretty cogent as to why people throughout history want to enact socialism based systems and why they eventually fail.

> capitalism still kills millions of people a year with starvation and preventable diseases

as if somehow, it's the responsibility of those capitalists to ensure those starving people have food and diseases cured - for nothing in return.


Why else do we even bother making a society?

To me it's equally as wild that you say such a thing when no system in human history did as much as capitalism to alleviate hunger and disease. In fact all other systems combined still can't touch the progress we've made to eradicate famine and disease while "under capitalism".

You truly do get the best and the worst with dictators, as history has shown.

Considering the state of current democracies, I don't think I could trust my peers to make sound financial decisions at an organizational level collectively. It's different from a democracy in government because I can, without much friction, leave a company at any time if I feel that the company is not doing well, not so with a nation-state.

I personally have started and run a coop, but that worked because we were all friends who wanted to collectively pool our resources to get new clients as a software agency, if it were a larger cooperative, I would agree with the sibling commenter that they generally do not work as well as top down corporations.


CEOs aren't there to serve the rank-and-file, they're put there to serve the board and investors.

And therein lies the problem.

and why is that a problem? The shareholders are the owners of the company. They (indirectly) dictate what the company ought to do; which is to make profits for them.

You want to take control away from the board and give it to employees? https://ai-ceo.org has a plugin for that!

Have a look at what Mondragon does in the Basque country for a cooperative of coopertives.

One of my favorite attributes about left-wing activism is that the solution to everything is to overthrow the system. It's like the whole "we have tech debt; we need to rewrite everything". Zero threat.

You don't need to overthrow anything to have worker owned companies, they exist right now.

It's called a union. If workers unite and refuse to suddenly supply their labour it'll cause a lot of changes fast, and sometimes the capitalists just leave and the workers can appropriate the company.

Though it is often a good idea to start small and get a little win, and then wait and recruit, and then push for another small change, and repeat. That way the union establishes credibility, solid membership and a reasonably sized war chest for things like getting rid of a CEO, which will take more time and effort than for example protective gear or longer pauses in the work.


My dad had a manager recently that he started calling "VPGPT" behind his back.

He felt like the guy primarily spoke in truisms and with long-winded statements that ultimately said very little but gave the illusion of having depth. He wasn't a huge fan.


I had one of those. I thought he was fooling upper management until I started working directly with the CTO, who was very open about the fact that this VP was more about performance than execution. He explained that the guy’s overwhelming desire to look good and be viewed as a thought leader also made him very receptive to incentives. You could give him a task, tell him it was important for his career, and he’d get it done because he was so afraid of looking bad. If it was something he could write about on LinkedIn or his personal newsletter, even better.

The challenge was setting the incentives correctly so he couldn’t game them. He was tasked with hiring a team to accomplish some specific goal once and the first set of people he tried to hire were not qualified at all, they were just the first people he could get to interview and want to join. You had to watch him like a hawk to make sure he wasn’t running yet another performative game.

The other weird part is that a lot of candidates were really impressed by him. He dressed well and had hair that obviously took a long time to do every morning and he spoke with confidence. This was remarkably effective at convincing many candidates that he knew what he was doing and could run a tight ship. Lot of disappointed people trying to leave his team after the first 6 months.


They tie up heavy, cumbersome loads and put them on other people's shoulders, but they themselves are not willing to lift a finger to move them. Everything they do is done for people to see: They make their desks wide and the ties on their garments long; they love the place of honor at corporate retreats and the most important seats in the interviews; they love to be greeted with respect in the meeting rooms and to be called a Leader by others.

...the more things change, the more they stay the same...


Matthew 23:4, nice.

“Meet the new boss, same as the old boss”

1971 was such a good year for music (see: https://en.wikipedia.org/wiki/1971_in_music and Apple TV+ came out with https://www.apple.com/tv-pr/originals/1971-the-year-that-mus... a few years ago)

> the guy primarily spoke in truisms and with long-winded statements that ultimately said very little but gave the illusion of having depth.

The term for that is deepety.

https://rationalwiki.org/wiki/Deepity


It used to be the case that speaking in that register was a strong signal of a social pedigree and an expensive education. Now it's a signal of people being brainless and easily replaceable.

You're also seeing it in politics where politicians that don't think before they speak too much are seen as more authentic and competent than ones that hire teams of consultants to carefully massage every statement they make.


> It used to be the case that speaking in that register was a strong signal of a social pedigree and an expensive education.

It's not the same register. There's a vast difference between the extemporaneous speech of the old school patricians and today's C-suite corp-speak.

Here's Robert F Kennedy Jr. speaking off the cuff at a public event moments after hearing that MLK had been assassinated: https://www.youtube.com/watch?v=A2kWIa8wSC0&t=238s.

That's the speaking style of social pedigree and an expensive education. You'll tend to find it contains lifts of poetry, like Kennedy's evocation of Aeschylus, and little reference to 'synergy', 'collaboration', 'efficiency', etc.


Maybe our standards are rock bottom now, but that speech did not seem to me to be "off the cuff." It seemed rehearsed, polished, edited and wordsmithed, like any other prepared speech by a politician. If he was truly winging it, then wow, what a talent!

This effect will probably be more pronounced in coming years. Being able to speak and act in ways that are obviously not AI.

The thing is that AIs are also good at mimicking informal/off-the-cuff speech, too. And people are lazy, so they'll fall back on "casual" cliches, which AIs (and consultant-driven politicians) will be just as good at mimicking. It's going to be the rare person who can stay ahead of that. It'll be an endless treadmill.

Unfortunately our CEO at a 5000 person company that heavily uses Slack, still hasn't activated their Slack account, so I don't have much to finetune the bot on.

I have gained respect for your CEO. Good for them.

Yup, no reason to sully yourself communicating with the rabble

It would be one thing if their employees didn't have to use it...

I'm saddened to tell you fellow hackers here that the real world will be more like: "Replace your knowledge worker with an LLM" ordered by your CEO.

I feel most bad for the other frogs in roles whose pots haven't even been put on the AI stovetop yet. I have general feeling they feel they're safe.

Looking at you specifically, Project/Program managers with little output other than attending meetings and asking for status updates and futzing with JIRA to get reports.

The dev frogs might not even be finished boiling before you notice your steam.


Yes, due to the power of the relative positions, but I'd like to point out

>Replace Your CEO with an LLM

Lossless, as they're already constrained so tightly in their functions by regulations.

>Replace your knowledge worker with an LLM

Lossy, in many cases.


This makes perfect sense when you consider who is making the decisions. If you are trying to, say, replace an ad-buying team with an AI that supposedly does the job better, your goal has to be to sell it at a high enough level that the political pressure of having the old department employed doesn't tank everything regardless of results.

One can try to sell to a CEO than an IT department is a cost center, and they can bring in the same results for less. But to replace a CEO you need to sell to the board, and the board is typically full of CEOs. Even a far superior CEO that costs $1 a day is a hard sell when the CEO is your buddy, and you aren't going to be making a mint as a board member for getting rid of said buddy.


If far superior AI CEOs existed then only one board would need to replace their CEO with AI. The company would then dominate the marketplace because of the far superior decisions being made, and the rest would be forced to follow.

If AI CEOs were slightly inferior to human CEOs, but cost $1 per year, it still wouldn't be worth replacing an expensive human CEO, unless that CEO's compensation was worth more than a few percent of revenue (it never is).

I wouldn't want to invest my money in a company which was making worse decisions to try to save money on the CEO, when the money paid to the CEO is going to have very little effect on my return on investment. That sounds like a good way to for my investment to underperform. The only winner in that scenario is online forum posters who want to feel jealous of the abstract concept of a "CEO".


In my past I was the technical co-founder for a successful startup and eventual exit. I had a background in marketing automation and analytics.

At different points I acted as CTO, CEO and COO and ended up at a CTO but with a cutout that I ran the "renevue" team.

Except for glad handing VCs in the early stages the first thing that should be automated away is most roles in the C-Suite.

You'll still need a flesh and blood figurehead for external meetings as well.

Many software engineers seem to be extremely blind to practical concerns around positioning, market fit, pricing, etc. and may bring certain idealogical biases and "purities" into their decision making that sabotage their success, just put yourself in the hands of AI and trust it.

The simple fact is most C-suite teams don't perform better than the macro environment they exist in, so getting "decent" AI decisions in place and then energetically executing you'll beat most teams, and in early stage startups.

If some people say we can replace the bottom 20% of devs right now, we can probably safely replace the bottom 40% of c-suite roles.


Are they blind or do others get paid to do these jobs? I've seen it too many times, no good deed goes unpunished.

Ridiculous. No CEO worth replacing is on slack - this needs to be trained on, and run on teams.

The OP is clearly just a fun tech demo to show off their platform, they're not seriously advocating to replace CEOs with LLMs.

> All the components—scraping, fine-tuning, inference and slack event handlers run on Modal, and the code itself is open-source and available here.

Does anyone know how much it would cost to put this together? How much does it cost to fine tune Lama 3.1 on someone's Slack messages?

I'm guessing it depends on the number of messages, but roughly for 10k-100k messages, at 200 characters each (guessing about 50 tokens each, so 500k-5M tokens)?


I'm mostly curious about the decision to do fine-tuning here instead of synthesizing a good system message, maybe a RAG setup with access to the slack database, etc. I tried fine-tuning on a content generation case when GPT 4 first came out and we had much better results without it. Also, you don't get the benefit of upgrading to new models when they are released.

I worked on this project a bit! These are great questions.

Fine-tuning works very well for style (as opposed to factual knowledge), which is all we're trying to achieve here -- as another commenter put it, it's a "comedy bit" for a company Slack.

In fact, fine-tuning for style works well enough that we find it pretty easy to just re-train when new models come out. There's sometimes some YAML-fiddling required to get training frameworks to work with different model series (e.g. a DeepSeek series model vs a LLaMA series model), but it's not too onerous. IMO, the ideal ML pipeline looks less like the bespoke process common these days and more like a materialized view (shouts to pgML). That's easy to automate and so reduces the gap with prompting.

On the other hand, it seemed harder to craft a generic system prompt and a generic retrieval system that would work across organizations to define user communication style.


Thanks for the reply! Makes sense. I will definitely keep this in mind. Much appreciated.

I think you're putting way too much thought into what is basically a comedy bit.

Please https://ai-ceo.org is the only real deal in that field.

I think they're burying the lede here, they released a DoppleBot you can train on Linus's messages from the lkml as the ultimate arbitrator of coding disputes.

Surely, it simply accuses everyone (including old and well-known contributors) of being a Russian bot and, as an explanation for this wild accusation, merely states that he is a Finn and knows history?

So, all this effort to enact a joke about the emptiness of bad leadership?

We are in the golden age of pointless AI!


Unrelated, but I wonder if there is a test to see how many of the commenters are bots vs actual humans

Could we use it to replace our CTO when the sales team asks feasibility questions ?

Or your government. See the TV show Travelers for an example.

Right, except who's going to be held accountable?

Rather than CEOs, why not mid level managers?

Yet again HN gets weird with highly contradictory popular posts trending simultaneously

https://news.ycombinator.com/item?id=42923870 "A computer can never be held accountable"


I'm unclear what the contradiction is? When was the last time a CEO was held accountable, for any value of "accountable" that doesn't involve repeatedly taking golden parachutes and failing upward?

usa could do better but they get sentenced regularly https://search.brave.com/search?q=ceo+convicted+site%3Ajusti...

I mean, this is clearly an unserious post. This is more of a fun tech demo than anything.

imagine the share price jump when the first corp announces this!

Maybe some good will come from this tech afterall...



Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: