| Jan 04 |
The best way to learn about agentic coding is to install one and painfully prompt your way into your first project.
Pain reducers: give it access to your full (backed-up) codebase and have it do a security scan. Or have it generate a few tests.
Just get used to this tech. Youāll have to sooner or later.
|
13 |
0 |
7 |
832 |
306 |
. |
| Jan 04 |
Ever since I started coding with agents instead of typing every line, coding has been fun again.
Particularly as a solopreneur, getting into deep flow is costly and often zaps energy from other tasks.
Watching Claude Code do that work and diving into the final PR is bliss. š§āš³š
|
69 |
4 |
25 |
3.7k |
280 |
. |
| Jan 03 |
I, too, would love to be a spontaneous foreign policy expert.
Oddly, looking at my feeds, the people from back in my PoliSci university days, the ones I took foreign policy courses with, are not the ones tweeting ridiculous opinions.
Isnāt that something.
|
17 |
0 |
1 |
2.7k |
258 |
. |
| Jan 03 |
Don't just say "fix it. make no mistakes."
Instead, give it a process and milestones. Context + status-quo + reflection + resolution.
|
15 |
0 |
2 |
2.3k |
134 |
. |
| Jan 03 |
Getting feedback from my readers is SUCH a joy! š„° https://t.co/gTNN8Mbngn
|
15 |
0 |
3 |
2k |
73 |
. |
| Jan 03 |
Know-your-customer in one image.
|
21 |
0 |
7 |
3.8k |
32 |
. |
| Jan 02 |
I am realizing that my episode about the Dead Internet theory (and the onslaught of too many machines talking to machines) has the episode number of the 429 HTTP status code.
429: Too many Requests
Couldn't be a better fit if I tried.
|
20 |
0 |
5 |
3.9k |
236 |
. |
| Jan 02 |
It's not looking too good: https://t.co/minX4BINez
|
5 |
0 |
1 |
3.5k |
50 |
. |
| Jan 02 |
"Everyone knows it's a bubble" is the biggest bubble.
|
28 |
0 |
18 |
2.9k |
53 |
. |
| Jan 01 |
Tweets with this image always go viral š«¤
|
8 |
0 |
5 |
2.6k |
40 |
. |
| Jan 01 |
Weāre now in Q2 of this century.
|
17 |
1 |
5 |
1.6k |
32 |
. |
| Jan 01 |
https://t.co/BIWjUibyXI
|
4 |
0 |
1 |
2.8k |
23 |
. |
| Jan 01 |
AI is the ultimate gatekeeper-breaker.
Don't understand a complicated error message? AI will explain it for free.
Don't know what that corporate email really means? How to say "no" courteously? Understand those rental contract clauses? AI helps.
Some people REALLY hate this.
|
64 |
6 |
33 |
4.7k |
278 |
. |
| Dec 31 |
Solid damage control.
|
2 |
0 |
0 |
1.3k |
21 |
. |
| Dec 31 |
haha, yeah, right https://t.co/cvNbMHAGDF
|
4 |
0 |
4 |
2.4k |
41 |
. |
| Dec 31 |
My wish for 2026 is for Claude Code to turn
"āŗ Now let me examine the XYZ model and related data structures:
āŗ Let me also examine the navigation structure and routes:"
into
"āŗ Successfully loaded my full understanding of your codebase from the 315 prior Claude runs."
|
37 |
0 |
6 |
2.9k |
274 |
. |
| Dec 31 |
My year in review: pretty amazing.
|
31 |
0 |
10 |
2.5k |
34 |
. |
| Dec 30 |
Of course, you can't rely on agentic code systems to write meaningful tests completely automatically. They've got much better at it, but you still need to read those tests to understand if the thing being tested is actually the thing you want to be tested.
But beyond that, access to testing has become so easy that now feels like I would just automatically do it on every new project because it is so easy. Just this tiny little improvement has significantly increased the quality of the software that I produce.
|
9 |
0 |
1 |
1.4k |
514 |
. |
| Dec 30 |
Before LLMs, I didn't test. I "had the full picture" (until a bug broke prod, of course.)
Because of LLMs, I learned how to test. I had it write tests for me, then fixed them to make sense.
You think LLMs write bad code? You should have seen mine before LLMs.
Much better now. https://t.co/TRGzAk2mTj
|
64 |
6 |
17 |
4.8k |
303 |
. |
| Dec 30 |
And here I thought hallucinating was an AI problemā¦
Good to see that gaslighting is still very much a human skill.
|
13 |
2 |
3 |
2.7k |
115 |
. |
| Dec 28 |
Nerd drama is the best.
|
5 |
0 |
2 |
2.9k |
23 |
. |
| Dec 26 |
Also, Europe: https://t.co/eXTfCB9mWF
|
27 |
1 |
6 |
2.7k |
37 |
. |
| Dec 26 |
You'll find the podcast episode for this at https://t.co/snxDFJ09H7
|
1 |
0 |
2 |
1.3k |
67 |
. |
| Dec 26 |
Please feel free to subscribe to the newsletter at https://t.co/c4fJ5q9Hno. Once a week, deep insights into a founder-relevant topic.
|
0 |
0 |
1 |
1.3k |
133 |
. |
| Dec 26 |
Everyone is celebrating Christmas, so who keeps engaging? š
Today's TBF newsletter issue is all about the "Dead Internet" theory and how we, as SaaS founders and technologists, can avoid contributing to a world where only machines talk to machines.
https://t.co/nkj6cQVN8o https://t.co/L122c4oqKc
|
6 |
1 |
8 |
2.3k |
298 |
. |
| Dec 25 |
Assuming an image is AI will be the new assuming an image is not AI.
|
11 |
0 |
9 |
2.6k |
68 |
. |
| Dec 24 |
True sales enablement.
|
10 |
0 |
3 |
3.6k |
22 |
. |
| Dec 24 |
Merry Christmas š„°
|
44 |
0 |
16 |
2.3k |
17 |
. |
| Dec 22 |
So glad I could contribute to their business success by being in the 27th percentile for product usage! š„°
|
13 |
0 |
3 |
1.8k |
105 |
. |
| Dec 22 |
Hear me out: I don't want every single service I use to do a "2025 wrapped", showing me just how much they monitor every single one of my in-app activities and how they compare to my lookalike customers.
|
30 |
0 |
8 |
2.3k |
203 |
. |
| Dec 22 |
The article about the fire is here: https://t.co/Wrj1BoTr5O
My fav song is this one :) https://t.co/eIOxKEPENC
|
2 |
0 |
0 |
1.5k |
111 |
. |
| Dec 22 |
Bus factors don't just exist for bootstrappers.
VOLA, an (objectively great prog) metal band, had their touring equipment ($200k+) go up in flames when a storage facility fire broke out. Insurance will pay $30k (lol), everything is gone.
So far, fans have pledged $70k in their GoFundMe.
Looks like they'll be able to recover from their fans alone. I've seen some fans donate thousands of dollars. True fans, you could call them, in Kevin Kelly's vernacular.
Between merch sales, tickets, and these donations, they'll find their way back into profitability. The lesson for me as a founder is that disaster strikes whether we like it or not, but recovery depends on a strong presence with the people who care about your work. Engaged supporters MATTER.
Also, you should really listen to VOLA.š¤
|
11 |
0 |
5 |
2k |
798 |
. |
| Dec 22 |
Haha. the race is on :D
|
23 |
1 |
6 |
4.2k |
23 |
. |
| Dec 21 |
I wonder when someone's going to release `left-pad` as an agentic "skill."
|
10 |
0 |
1 |
2.2k |
74 |
. |
| Dec 20 |
Any field of expertise eventually reveals this truth: it is not stirring the sauce that is complicated and takes years of experience; it is knowing the right ingredients, pots, spoons, temperatures, and precise steps that make it hard.
Same goes for coding, writing, public speaking.
The true value lies in an expert's ability to deploy their taste to judge quality and to discern good work from bad work at a glance.
|
12 |
0 |
2 |
4.3k |
419 |
. |
| Dec 20 |
So are we still doing MCPs? Or is it skills now?
Been thinking of making Podscan more "in-AI-usable." Right now, it's a REST API with all the trimmings, but I wonder how I can make it more pluggable.
|
18 |
1 |
9 |
4.8k |
200 |
. |
| Dec 19 |
Machines changing their point of view when presented with evidence is my favorite thing. Particularly because people have such a hard time doing that.
āStrong opinions weakly heldā is the LLM way.
|
13 |
0 |
3 |
2.6k |
197 |
. |
| Dec 19 |
This is a repeat of what happened when computers and digital case lookups were introduced to the legal industry.
Anyone who preferred flipping through tomes in libraries was quickly left behind.
Back then, lookup and retrieval were revolutionized.
Now, it's reasoning.
|
8 |
1 |
1 |
2.4k |
271 |
. |
| Dec 19 |
The moment we'll run out of Science Fiction stories to tell, I'll start believing in the Singularity.
|
10 |
0 |
6 |
1.5k |
101 |
. |
| Dec 19 |
And THIS is why we need generative AI :D
|
6 |
0 |
1 |
2.6k |
40 |
. |
| Dec 19 |
This is the kind of stuff you can expect enterprise companies to already be tinkering with in their software engineering labs, trying to find a way away from the AI platforms hosted and controlled by somebody else, particularly in industries where regulation prevents most cloud software usage.
This is the kind of work that will facilitate adoption of AI tooling into these companies, and it becomes critical for software companies to consider that not only do they have to do on-premise deployment of their software stack, they also have to be able to integrate into the on-premise deployments of their AI clusters.
|
3 |
0 |
2 |
838 |
619 |
. |
| Dec 19 |
4 Mac Studios connected via Exo using the new RDMA thingy.
Consumer hardware (albeit still expensive), sizable LLMs, 1.5TB of VRAM. 30+ tok/s on DeepSeek 3.1 (the 671B model).
Bit expensive to vibecode, but completely private, permanent, customizable.
https://t.co/aF7zJnuBuA
|
8 |
2 |
3 |
2.9k |
278 |
. |
| Dec 19 |
Kinda weird that every other person in my parentsā generation warned me not to trust anything I saw online when I was a kid in the late 90s.
The same people now consider this to be an amazing and totally real thing, without a flicker of a doubt. https://t.co/qiBKuoEk5V
|
8 |
0 |
6 |
2k |
270 |
. |
| Dec 18 |
https://t.co/WRkrXjLAdc
|
7 |
0 |
2 |
2k |
23 |
. |
| Dec 18 |
Remember that the ONE thing your churned customers will remember is how easy it was to part amicably.
If you want to fuck this up, I recommend:
- call-to-cancel
- no-refund policy
- dark patterns in every interaction
- vendor lock in
(Serious note: Olly is doing things right)
|
13 |
0 |
3 |
2.1k |
280 |
. |
| Dec 18 |
Sometimes I watch the map of people currently using Podscan, and I'm mesmerized by the fact that this project is being used by professionals all over the world. https://t.co/HvoyIH7E4y
|
31 |
0 |
3 |
1.9k |
184 |
. |
| Dec 18 |
Allowing non-technical people to create AI slop has been the most impactful form of technology democratization since self-publishing enabled everyone to share their precious thoughts.
|
26 |
2 |
6 |
1.6k |
183 |
. |
| Dec 18 |
In 2026, weāll see a mainstream shift in virtual agentic employees.
Paying for an āautonomous marketing agentā with full access to your blog, email, and socials will turn from weird to useful.
Crafty founders already build and run these things internally. And they convert.
|
14 |
0 |
8 |
1.9k |
275 |
. |
| Dec 17 |
I see more and more paying users using Podscan to track emerging trends and investment opportunities in podcast conversations for venture funds and financial institutions.
Probably nothing.
|
46 |
0 |
7 |
3.7k |
190 |
. |
| Dec 17 |
A merchant's refund policy stood firm for ten thousand customers.
One whispered "chargeback."
It had never existed.
|
13 |
0 |
5 |
3.5k |
119 |
. |
| Dec 17 |
Considering everything AI-generated to be "AI slop" is actively holding back people who could leverage these powerful tools to build amazing things.
|
33 |
3 |
21 |
2.8k |
148 |
. |
| Dec 17 |
https://t.co/wUeePMtHZr
|
2 |
0 |
3 |
2.7k |
23 |
. |
| Dec 17 |
Bet they had really solid conversions on that one.
|
6 |
0 |
2 |
4k |
50 |
. |
| Dec 16 |
Yeah, which chronoton fluctuation during warp travel did get us into this timeline...
|
10 |
0 |
1 |
2.3k |
85 |
. |
| Dec 16 |
This is the feature I think edge models (the on-device stuff thatās getting better and better) will be perfect for. Post-commit hooks triggering a background code review that informs you of potential issues. All local, offline, and secure.
|
29 |
2 |
8 |
4.5k |
239 |
. |
| Dec 15 |
Great pre-filtering effort when recruiting devs. Probably won't fool the AI-powered resume submission bots for too long, but even then, it will significantly reduce the spam.
I wonder if similar skill checks can be used in non-dev industries.
|
19 |
0 |
4 |
3.8k |
243 |
. |
| Dec 15 |
This used to happen once a year. Then once a month. Now, we can see this almost every week.
Anyone good at extrapolating?
|
11 |
1 |
4 |
3.9k |
122 |
. |
| Dec 15 |
Pull requests and GitHub/JIRA issues are human process markers that happen to relate to code. Agentic tools can handle these fine, but they likely can benefit from a more source-proximate format.
|
8 |
0 |
2 |
1.4k |
195 |
. |
| Dec 15 |
I was thinking about this a lot recently when I imagined the ānext-genā coding language/framework/env.
Prompts are architecture and intent documents. They need to become part of the code, as a trace/comment/metadata. Whatever the AI equivalent of a āgit blameā result is, too.
|
15 |
1 |
7 |
3k |
277 |
. |
| Dec 13 |
And here we have a small inkling of whatās to come in fundamental computer literacy.
Folders, files, what does it matter when all content is consumed in-app anyway?
|
16 |
0 |
7 |
4.7k |
165 |
. |
| Dec 13 |
Grinds my gears when people call themselves āoverachievers.ā
This only ever works with someone elseās expectations. Itās always external.
No one who followed their own dreams ever overachieved.
|
22 |
0 |
13 |
2.1k |
196 |
. |
| Dec 13 |
Some people could be offered a teleporter beam onto the Enterprise and would complain about having to stand still for a few seconds.
Dan is spot on: people getting grumpy over price when facing a miracle is so tedious.
|
21 |
0 |
6 |
4.3k |
219 |
. |
| Dec 13 |
Couldnāt be happier with the feedback Iām getting for my latest podcast episode š„°
|
11 |
0 |
0 |
2.5k |
81 |
. |
| Dec 12 |
Do you send out cold emails for your SaaS?
|
9 |
0 |
6 |
2.9k |
42 |
. |
| Dec 12 |
And, from another angle, it shows the taste baked into the models that were used for the job. Something you either like or don't. Corrective prompting might allow you some flexibility, but every LLM has the consequent taste of its training data.
|
5 |
0 |
3 |
1.4k |
245 |
. |
| Dec 12 |
Generative AI is a necessary milestone on the way to a functioning holodeck.
And like most things, Star Trek even thought about intellectual property rights in Voyagerās āAuthor Author.ā
That and derivative slop. And deep fakes.
In 2001! Way ahead of its time. https://t.co/aG25sXdPWE
|
10 |
0 |
2 |
1.7k |
288 |
. |
| Dec 12 |
A quick functionality summary: It finds people who most recently signed up for a trial and were scored as good prospects for my ICPs.
It looks for similar companies, similar size, and location as those people, and figures out the right people to approach at those companies.
It lets me review it in the email that was drafted to be sent to them. Pretty simple "hey, interested in this kind of stuff" email.
That happens several times a day and guarantees a constant flow of interesting outreach targets that fit exactly who's already using the platform. And the more people from one kind of industry join, the more of their peers I find and reach out to. Buzz begets buzz.
|
4 |
0 |
1 |
1.3k |
675 |
. |
| Dec 12 |
In a single day, I had my coding agent build an "existing customer lookalike and outreach tool" right into my Podscan admin panel.
And I realized that the one thing AI struggles with is reliable data. Never the code.
It can build complex things. But it can't build facts.
|
46 |
1 |
8 |
3k |
273 |
. |
| Dec 11 |
Just had a call with a prospect, and they were all excited about automations, integrations, and how to set up their Podscan keyword alerts so they cover not just a brand but the broader market they operate in.
I just spent half an hour on excited back-and-forth brainstorming and turned it into a highly well-structured article with great examples I instructed Claude to switch and "normalize" so that the customer's privacy can still be retained.
Not only did I help a customer, but I got another way of talking about their problems and the solutions that they might need, which is now codified in an article that will soon start ranking.
Low-effort, high-impact. And can be fully autoamted, too.
|
4 |
0 |
0 |
890 |
701 |
. |
| Dec 11 |
An extremely effective way to turn customer support conversations into high-impact blog posts (or KB articles) is to record customer calls, take the transcript, and have Claude/ChatGPT turn the conversation and its main problem(s) into helpful articles.
Works REALLY well.
|
23 |
0 |
9 |
1.8k |
273 |
. |
| Dec 10 |
Whew, my https://t.co/pUqGI7kwYe newsletter hasn't been sent for a couple days because of this strange postgres error.
Apparently, when null bytes are IN a record in the database, even a select query fails horribly.
Yikes. https://t.co/v7ElkqGUa7
|
20 |
0 |
3 |
2.6k |
249 |
. |
| Dec 10 |
Great thought piece on mortality as a requirement of progress (and the impetus to act). Recommend reading this 5-minute piece.
The quote made me think of why founders keep building even after their first exit. The journey is what matters.
https://t.co/sekBUgfBAR https://t.co/NyUSCobrW9
|
21 |
2 |
6 |
2k |
288 |
. |
| Dec 09 |
I hear Rust is a great programming language.
|
31 |
0 |
22 |
4.9k |
44 |
. |
| Dec 09 |
And now, @MistralAI throws an agentic CLI into the mix as well.
Can't wait for the meta agentic CLI, the one that runs your prompt on 5 different agentic agents using 5 different models in 5 git worktrees, scoring the "best" implementation candidate and suggesting it to you. https://t.co/oqleBgdVBK
|
41 |
2 |
10 |
3.9k |
301 |
. |
| Dec 09 |
Haha, this is wild.
I mean, a fine is just a tax with a gun. But I wasn't aware of the discrepancy.
It would be smart to grow the income tax portion of the equation. But I have a feeling that legislative efforts focus much more on the fine part.
|
9 |
0 |
6 |
3.9k |
248 |
. |
| Dec 09 |
āLLM seedingā success is a direct consequence of āconversational mindshare.ā
Itāll be much harder to trick data ingestion systems if their AI gatekeepers discard automated posts with no social proof.
A lonely story posted to a forum wonāt rank. That seed needs to grow first.
|
7 |
0 |
3 |
2.5k |
277 |
. |
| Dec 08 |
People put the weirdest things into their podcast's RSS feeds.
Seriously, all data collection at scale ends up being 99% data cleaning and validation. https://t.co/0o71lFxy1s
|
28 |
0 |
2 |
3.3k |
175 |
. |
| Dec 08 |
Devs: āAI sucks, itās non-deterministic.ā
So are people. Much more so, and much less intelligently so.
The challenge is not to enforce determinism, but to set boundaries that reign in that chaos.
AI might be hallucinating, but itās also very good at following clear rules.
|
68 |
1 |
32 |
4.3k |
277 |
. |
| Dec 07 |
Do you skip ads/sponsor reads on podcasts?
|
14 |
1 |
17 |
4.7k |
42 |
. |
| Dec 07 |
The biggest opportunity in AI is verification.
|
21 |
1 |
3 |
2.9k |
46 |
. |
| Dec 07 |
Any AI system can generate arbitrarily complex output. It will do so happily.
Verifying the factual integrity of that output is the actual challenge.
We get around by tasking a different AI to do a sanity check. But thatās also a gamble.
Solve this, and youāll have a unicorn.
|
11 |
1 |
3 |
1.6k |
280 |
. |
| Dec 05 |
And just to clarify, I fully agree with Jack because that has been my own experience as well.
Obviously I'm biased because I'm a solopreneur, but my throughput on the smaller tasks that would have taken me a day just to figure out is now a matter of minutes. All the potential contractor collaborators I could have paid won't be needed by me and likely many founders like myself.
Now amplify this for much bigger businesses and you see how significant that change is for people who are trying to get started with freelancing.
|
11 |
0 |
0 |
1.1k |
529 |
. |
| Dec 05 |
Not because it's impossible to use informational advantage to make moneyāevery entrepreneur does very much the same thing with their knowledge about the market and the businesses they create from it.
But assuming that betting on outcomes is going to level the imbalance or overabundance of information is a... BIG gamble.
|
2 |
0 |
2 |
834 |
322 |
. |
| Dec 05 |
"Prediction markets surface truth to people" is the inverse of every single past example of how humans act in chance scenarios: fixing games to win, insider trading, and market manipulation.
I'm VERY cautious with this not-quite-gambling-but--very-prone-to-fraud development.
|
11 |
0 |
4 |
3.1k |
276 |
. |
| Dec 05 |
Installing Claude Code just for having a brainstorming partner is absolutely worth it.
I often dictate prompts like this one, and it's super helpful. It dives into the code, highlights blind spots, and even estimates their relevance and criticality.
Beats blindly committing. https://t.co/0wixiD6pR3
|
70 |
1 |
23 |
4.2k |
301 |
. |
| Dec 05 |
Pantoneās color of the year is a grayish white.
Taking risks, I see. https://t.co/ifS0WCmTby
|
60 |
0 |
13 |
3.6k |
94 |
. |
| Dec 04 |
2 days left.
I'm running a time-limited Black Friday sale of The Bootstrapper's Bundle: all my books, all my courses, all formats, for $25 instead of $100+. Grab it here:
https://t.co/heHTsJuwT9
|
15 |
3 |
4 |
2.1k |
198 |
. |
| Dec 04 |
In the world where all these permissively licensed code fragments are used to power combat drones and military vehicles, maybe it is a good thing that we're exploring a more restrictive kind of licensing for the software we write?
|
21 |
0 |
3 |
3.5k |
230 |
. |
| Dec 04 |
Me, a successful angel investor: https://t.co/e6uMda4moS
|
71 |
0 |
16 |
4.5k |
56 |
. |
| Dec 04 |
Love how there is an "exit" sign very strategically placed in this pic š¤£
|
12 |
1 |
3 |
4.2k |
72 |
. |
| Dec 04 |
You come here for lukewarm takes on AI, SaaS, and MRR, but how about something much more important: Unreal Tournament 2004 is coming back:
On Linux and macOS, too! :D
https://t.co/c9xTKujieX
|
24 |
0 |
8 |
2.3k |
193 |
. |
| Dec 04 |
From my Spotify for Creators wrapped.
This numbers might not look massive, but every single person behind these figures means the world to me š„° https://t.co/7VBFOkFmq3
|
40 |
0 |
9 |
2.9k |
169 |
. |
| Dec 03 |
Every day, more people are getting confused and asking, "Is this AI?"
At first, it was the glaring AI-generated image. And a slop blog post.
Now, it's any well-written reply that contains, god forbid, an emdash.
Soon, it'll be any cohesive thought that isn't obviously biased.
|
45 |
1 |
22 |
4.1k |
279 |
. |
| Dec 03 |
What could possibly go wrong? https://t.co/pHBZMErju0
|
34 |
0 |
8 |
3k |
53 |
. |
| Dec 02 |
And just like this, dev tools have a completely new acquisition and monetization vector.
|
28 |
1 |
2 |
3.8k |
88 |
. |
| Dec 02 |
Something I see a lot with my more corporate customers: they trial Podscan on a personal account and only when theyāre ready to purchase, do they sign up with their business domain.
Thatās why every trial should be treated like theyāre a Fortune 500
In disguise :)
|
55 |
0 |
14 |
3.5k |
265 |
. |
| Dec 02 |
Podscan found a mention of Podscan in the 11th minute of @startupspod 9min after the episode was released.
I'll never tire of being amazed by this tech. It finds mentions before anyone even had the chance to listen to the whole show.
(Unless you listen on 2x like @robwalling)š¤£ https://t.co/gP1tjdgpnU
|
34 |
1 |
5 |
2.9k |
303 |
. |
| Dec 01 |
I have, I do, and I will keep doing it. Shows good results every year.
|
4 |
0 |
3 |
1k |
70 |
. |
| Dec 01 |
Are you running any Black Friday deals for your SaaS?
Why (not)?
|
9 |
1 |
12 |
3.9k |
65 |
. |
| Dec 01 |
As a developer, you will not be able to avoid being compared to this tool in output, development speed, and ease. Either you embrace it, or you have a really good answer as to why you wouldn't.
|
8 |
2 |
1 |
1k |
193 |
. |
