Can't make money online?
Teach people how to make money online.
Can't make money teaching people how to make money online?
Start an agency teaching people how to teach people how to make money online.
Can't get clients for your agency?
Sell a course on how to get clients for an agency.
11
If you want to delete your API keys on Fireworks, you'll need to do it one by one, then copy/paste the API key name.. over and over again.
Thank god for Claude Chrome!
Fireworks, the most highly-valued out of a growing group of inference providers, told customers including Harvey that it had been hit with a hack that exposed customer data. Some customers are now considering moving their business off the platform.
theinformation.com/articles/…
1
42
Don't feel bad when you see a startup's launch video that gets millions of views.
1. It was paid for
2. It doesn't correlate to PMF
ok so here's how we get to building a 100M company in two weeks before thanksgiving..
first, our preseed plan. we're going to raise $2M. why $2M? simple. $200k for the launch video, $200k to promote the launch video to xM impressions - i have six agencies lined up for this don't sweat with no paid partnerships tag duhh, and $500k for the tokens i'm going to obliterate to have a resemblance of a product. again its all token spend and no employee spend because vcs dont want that. solo founders is hot again so i can fully get away with this lfg.
ok, so next we use the token spend that i can vibe code and sell to "wink friendly wink" customers and have them give me LOIs and cARR - in return for using their products obviously. scratch mine, scratch yours. ok then i can fully wrap in the most creative financial metrics possible to show xM in revenue of ARR because who even knows what that means anymore.
and then i'm going to run the tighteeest possible process as possible for my "already oversubscribed seed". get some sick angels at a discount duhh, fomo VCs, get that tier 1 TS, shop it, and tranche it to the moon baby. first at 40M, then 80M and then finally a few angels at $100M.
of course i'm going to incorporate the company as late as possible even though all the machinery is there. so i can claim 0->10M in revenue on day 14 or whatever.
and yeah completely forgot that i need a sick vision and a full deck, that will need $100k for someone to drill that to me cause i ain't got no clue what this category even means. don't sweat, there's a fundraising expert who might take x% and some cash or whatever to help me whisper to the right networks. yeah, that's a thing.
diligence? who needs charts and due diligence, we live in the token economy, and lfg economy baby. elections and the market selloff? don't sweat, the seed market is still red hot, it might take a day more, but this will still most def work.
anyway, i'm going to go try this and i'll see you in two weeks eh?
(before i get obliterated this is satire, but every sentence in there individually has personally happened to me)
1
2
52
I actually cannot name one "ism" that Opus 5.5 has other than it saying "Fair" when you raise an objection.
They solved it?
24
In 6mo no one is going to have to read or write code anymore.
Or at least, companies will realize it.
Code review feels a bit like a relic of the past. Sure I still think I still think humans should have a hand on the wheel. I just no longer think thats as effective as good integration tests and observability so you can catch anything you missed.
1
2
268
Is the RSI in the room right now..?
SITUATION DETECTED: Bloomington, Indiana's city council voted unanimously this week to ban any recursively self-improving entity capable of greatly outperforming all humans on all cognitive tasks from inside city limits, citing existential hazards and loss of human control.
34
I wouldn't try to build anything on the harness level anymore.
It's not worth it, plus you're signing yourself up to a lifetime of bugfixing because the harness is incredibly complicated.
The models will eat up the harness over the next few years.
Wouldn't be surprised if harnesses even collapse further as AI becomes more intelligent.
We may not be so far from everything being ran on the webUI versions of Claude/Chatgpt nullifying any chance any competitors have.
Yep, source: tried to build one myself
The truth is, it may have been plausible to outship Anthropic/OpenAI in early Jan, but now you're at a lopsided disadvantage, even if you raised 100m+ you don't own the models.
The models are the product, and the harness has been eaten alive.
I've since stopped developing on it and just started using Claude Desktop again.
As models became more intelligent, I didn't suspect multi-agent orchestration would die because of it, but it's basically not a problem you can hope to solve unless you have unlimited tokens.
It was a really difficult product to build when I started (GPT 5.4 was SOTA at the time, and was still terrible) but considering there are only 2 major players and everyone pays for tokens via a subscription, it just doesn't make sense anymore.
2
1
194
Yep, source: tried to build one myself
The truth is, it may have been plausible to outship Anthropic/OpenAI in early Jan, but now you're at a lopsided disadvantage, even if you raised 100m+ you don't own the models.
The models are the product, and the harness has been eaten alive.
I've since stopped developing on it and just started using Claude Desktop again.
As models became more intelligent, I didn't suspect multi-agent orchestration would die because of it, but it's basically not a problem you can hope to solve unless you have unlimited tokens.
It was a really difficult product to build when I started (GPT 5.4 was SOTA at the time, and was still terrible) but considering there are only 2 major players and everyone pays for tokens via a subscription, it just doesn't make sense anymore.
I wanna be real here.
"GTM" basically doesn't mean anything unless you already have distribution.
You can't GTM unless people already know about you.
Otherwise, no one will respond because they don't know you.
Calling it "Cold outbound" is highly misleading.
41
I'll do you a favor: A decent backlink in 2026 likely costs $3-500. A single one!
DO NOT PURCHASE @JanuBuilds Effortless backlinks service.
Back in October I purchased the $200 package to get my SaaS, @projectecho_io listed on 100 directories. I did this to help boost the authority of my new SaaS and to save myself the time. I regret that decision.
Today I got an email from "Miguel" telling me that the work was done and I could review the report.
I went through the list of directories where my SaaS was listed and was surprised to see the following places listed:
1. Blogger
2. Medium
3. Substack
4. Pinterest
After clicking on the "proof links", I was shocked to find that new accounts were created under my name on Medium and Substack.
I have personal accounts on these platforms where I've published over 100 pieces of content and have an audience of almost 1,000 readers. I have mentioned Project Echo in more than a dozen posts on my Substack.
On Blogger, the post was some low quality guest post on what looks like a fake Blogger account.
These places were added to the list because of their "high DR" but of course search engines aren't taking that into account. This shows that Janu doesn't understand even the fundamentals of link building and SEO.
And the worst offense: The screenshot below shows another link building post for another client published on my new substack account. I've never heard of this brand and never gave any approval to publish this content under my name.
I can't comment much on the rest of the sites where Project Echo was listed, but that's beyond the point. I take my personal brand seriously and it's insanity for a service like this to take the steps they did.
I regret making this purchase and warn you before you do the same.
@JanuBuilds, you fucked up and hopefully this will be a learning experience so that you can make the necessary changes to your service, put better controls in place, and do better.
1
2
174
Yeah but the companies still want you to come in to "Show your expertise" of hand coding while stating that humans can't code well.
How much longer until the argument is dead?
I can confidently say that humans were not very good at writing secure infrastructure. The endstate will be hypersecure infrastructure and software for all, but we're going to go through some turbulence before we reach that.
43
Seeing all of these competitors undercut each other, did no one learn about "Race to the bottom" in school..?
1
53
The very first versions of the Facebook redesigns were exactly like this.
when you change something someone is familiar with it almost ALWAYS feels like a huge downgrade and there will always be some level of negative first reactions... but its basically impossible to make forwards progress without changing things. maybe i will add a way to use old version of the app but i'm not sure - also worth noting most people do not pay for the app! so i will listen more to the people who do.
1
74