Ben Skoog

[Ben Skoog @BenSkoog1]

@UnbeknownstDude @jamiequint It so funny because it should be extremely obvious that you cannot will consumer and consumer adjacent apps into existence. No amount of contrived hype can make it stick. Look at airchat.

I respect the effort and product quality though

[Ben Skoog @BenSkoog1]

@UnbeknownstDude @jamiequint https://t.co/SGtl7rHyZ1

[Ben Skoog @BenSkoog1]

the instinct hype is seeming increasingly VC astroturf

[Ben Skoog @BenSkoog1]

@pitdesi https://t.co/SGtl7rHyZ1

[Ben Skoog @BenSkoog1]

the instinct hype is seeming increasingly VC astroturf

[Ben Skoog @BenSkoog1]

@bitcloud It’s not weird though. It’s trying to be weird.

This is industry anointed neposlop and people are justifiably tired of it.

[Ben Skoog @BenSkoog1]

Obscure and jargony = easier to watermark?

[Kyle Mistele 🏴‍☠️ @0xblacklight]

what explains this why does it do this

[Ben Skoog @BenSkoog1]

@IJoukov @0xblacklight This is such a great point

[Ben Skoog @BenSkoog1]

@jappleby Fractional is just paying to get advice from someone who sees what a lot of companies do relative to their domain. They won’t own anything or actually do any meaningful amount of work.

[Ben Skoog @BenSkoog1]

@0xblacklight Is this actually happening? I have no problem understand Claude. Codex is great but Fable is still the best model.

[Ben Skoog @BenSkoog1]

Haven’t personally felt this with fable but I recommend listening to @dwarkesh_sp’s episode with @RyanGreenblatt.

He talks about how in the relatively near future agents will effectively communicate in neuralese and humans will simply not be able to keep up, which may enable them to conspire against us (intentionally or not)

The example here with fable is the just the very beginning.

Imagine models many times smarter than fable running on Cerebras like compute at 2500 tps.

That sort of system would run circles around our smartest humans.

Loss of control might come fast.

[Jaya Gupta @JayaGup10]

Fable 5 makes 0 sense for 99.9% of enterprises.

1.) It’s reaching a point where I’m dreading reading the outputs it generates.” The jargon and detours feel like a mathematician using non-standard notation that confuses people whose expertise doesn’t match the model’s.

2.) It has so much Claude-Speak. There are hermetically dense sentences, invented compound terms, metaphors that prioritize the model’s internal shorthand over human readability. People are piping the Fable outputs into Haiku just to translate it into plain English.

3.) Repeated reports of “first model that made me feel stupid,” “I can’t keep up,” “goes over my head,” and “intelligence delta” comparisons (high-IQ specialist talking past a normal person). Even strong engineers and researchers describe the writing as exhausting or unreadable. It is simply not accessible for people

4.) When Fable 5’s outputs require extra mental effort just to decode, you end up working harder to understand the AI than you would have spent solving the problem yourself.

5.) These problems compound in real workflows: in multi-agent setups, long sessions, or handoffs, the private shorthand and dense style mess up the outcomes. It makes the entire system / skill harder to supervise, debug, or trust.

[Ben Skoog @BenSkoog1]

yeah, you're definitely right on the PE holding period. I was thikning more around how the markets are probably just pricing in a terminal growth cap, which wasn't really what you were talking about anyway.

a separate thought: I think we are underweighting how much AI is going to collapse switching costs.

I do think we will see the Workday's of the world get their use cases and margins compressed and it will be easier to switch. you can only get away with "but you trust us and we'll take you out for steak dinners" for so long when competitors are offering something 10x better for a quarter the price.

[Ben Skoog @BenSkoog1]

@jasonlk @nrmehta @Fritzanity yes, and the more important question to ask given the speculation above is, do we really think companies born today will ever be using bloated system of record products like Workday?

no chance. so I don't think they are undervalued at all. happy to be wrong though.

[Ben Skoog @BenSkoog1]

@aporia9n This has been said 1000x on this platform and each time it’s somehow more unoriginal

[Ben Skoog @BenSkoog1]

@zeroxkyle Ah yes the lowly Palantir FDE in 2012

[Ben Skoog @BenSkoog1]

@pranavreddy https://t.co/SGtl7rHyZ1

Still waiting for one non VC to say something about this product lmao

[Ben Skoog @BenSkoog1]

the instinct hype is seeming increasingly VC astroturf

[Ben Skoog @BenSkoog1]

not getting the logical leap from "must be all channel" -> "must build all in house", but I think it depends on what you mean by support infra.

we basically built what you're describing as v1 of our product (moved into something adjacent because I came to a similar conclusion as you).

what I'll say is that you probably don't want to own high touch customer facing software, internal tools are another thing. it just becomes a headache and you'd rather be focusing on making your own product better.

[Ben Skoog @BenSkoog1]

@dakeenekid @BlendiByl I never get this with fable but did with the old versions of opus. Had to tell it to create a memory to stop doing it lol

[Ben Skoog @BenSkoog1]

@TurnerNovak @numeral @UseCorgi @FlexSuperApp @MonacoGTM this feels like an ad

[Ben Skoog @BenSkoog1]

building on easy mode:

- daily scan of Slack, Email, meeting transcripts, etc.
- extract customer requests, bugs, issues
- check against Linear or your ticketing system
- create new issues with links back to source
- autofix simple bugs

then:

- daily AI-human grooming of tickets + create sprints
- one-click build entire sprint
- orchestrator deploys sub agents for each ticket
- monitors sub agent progress
- handles conflicts, assembles a PR
- loops PR and watches for AI code review findings
- continues fix-watch loop until clean
- I merge

soon all you'll have to do is talk to people, have cool ideas and the world will simply rearrange itself around your intentions.

[Ben Skoog @BenSkoog1]

I don't like Opus 5's personality. it is patronizing and almost like it tries to be obstinate when it doesn't need to be.

one of the previous Opus versions was like this and I thought they had fixed it, but I guess not.

I really like Fable because it just wants to build, and to your point, maybe too much.

[Ben Skoog @BenSkoog1]

eh, this isn't the only option. there is a very real case for using AI to streamline buyer experience and procurement, making complex software much easier to buy.

vendor discoverability and matching is also improving rapidly - you can send an AI agent off to find the software that fits your use case best.

most companies will still buy if the the option fits and can be turned on fast.