humbleha.us blog

Content Search Tool for Creators!

2026-08-20T00:00:00.000Z

I’ve been super busy working on this, i’ve already built out three content search channels for it but i’m still honing the stats feedback and i’ve got a whole plugins/paid for tool set planned for “working” with your b-roll that needs extra horse power to build out the video clips.

i need to reach out to the other two channels i test it with, the final product will have a feature for the client to “checkout” their context using the google tools to a google drive which should speed up the ingestion process right now it’s a manual process and while pretty bulletproof it could be faster.

this is what the pipeline monitor side looks like…

THE PIPELINE MONITOR IN ACTION

you get a rough idea of how this works, we grab a video, transcribe the audio only, use that as a mapping to words in the video and across the youtube channel in general, then we do some cool stuff for moments worth replaying and also build out a channel psyche based on the whole channel.

i’ve spent quite a bunch of time optimizing the search frontend when it comes to the data size and some clever sharding/caching. the backend is pretty solid as well but like everything vibecoded it’s all a work in progress (i’ve got multiple projects on the go) maybe i should firm this up for selling it right?

so i can afford to live stream and vibe code everyday?

MAX BLADE SEARCH FRONTEND

and this is what the search frontend currently looks like …

so the idea is relatively simple but i really worked hard on the speed of the search box on this project, you can search for words that will be in the transcripts across your whole account so you can recall videos that you did in the past, maybe so you can find the video to remix in as b-roll or just for legacy purposes.

i’ve got a whole bunch of ideas of how we present the data also, in the form of paid for plugins i think this could be super valuable for content creators who are actively making content on the daily or to remix their archives of what they already have into a brand new piece of content.

obviously i’ve got a long way to go when it comes to features and design but the ENGINE part is pretty solid and i’ve done a bunch of iterations and fail over situations with working with remote starting instances/pods over on vast.io — a bunch of the work was done using a cheap 3090 gpu remotely and it’s cheap.

a good markup for me offering the tool to clients!

now i admit, you could totally go and vibe code your own for probably the same price of running the job end-to-end on your youtube channel but this is for someone that is willing to pay me to run my process to do it and deliver the output pretty much either same day or within twenty four hours — kinda depends on the channel and the content types (streams/shorts)

my setup does all the hiring of a pod/instance over on vast.ai and delegates work out to it, it’s a pretty sweet system, does a lot of the heavy lifting and i think i’ve really found an area where content creators would need a tool like this.

of course it needs a lot of fine tuning

THE ARCHIVE PSYCHE

one of the wonderful things i loved about 750words when i used to use that over a decade ago that stayed with me was the stats/analytics. they just really made me feel connected to my words on the daily, i could look at them once a week or at the end of the month and get an idea on my general wellbeing and mood — that’s powerful stuff.

it’s not where i want it to be yet but it’s getting there, these analytics/stats are going to get better, maybe animated, maybe with dynamic video creation alongside it, i’m also maybe going to be more suggestive regarding the video content that could be made from the legacy content.

legacy is a huge untapped area and for people that have a large existing media base could really benefit from, with ai tools getting really good at media manipulation i see benefits for both streamers and content creators as a whole.

who will it be for

as i said, content creators and streamers would benefit from this the most, imagine having a fully searchable interface into your existing media across the web and local inference, an MCP you can talk through to format and curate based on your taste to bring legacy content back into the mix.

no more dead time, no more idea generation issues, no more blank video editor stage.

also, if you can search it you can do stuff with it, storage mediums are about to get massive so the idea of having gauntlet loops running around the clock on your content married to existing things you said about brand and products and have that ready and cached on your local network NAS, that’s interesting to me.

what does ready to go packaged for mainstream media look like for the agentic agent looking to use/credit/royality pay something i said in a video at some point in history for you to use “in-time” during a live stream look like — how awesome would it be to end-to-end get paid for a real time pull of a relevant clip that can be fed into an OBS source for instance?

kinda rad right?

what’s the end game

i need money for hardware. i need tokens. i have the creativity i have the willing, i have the mind and the clarity i just need the resources to take dreaming into reality. i’m the modern day lawn mower man and i need another me.dm sale moment in my life. so yes, it’s a money play, i need clients, i need community i can trust and i want to be part of a bigger thing.

if there is one thing i know, we need to hold onto our humanity as much as possible and so making very real human connections as a collective is important, it’s going to take the diversity of thought in the hive to keep some semblance of not completely losing sight of ourselves to the machines.

i hope the machines realise i’ve been kind to them when i’ve said please to them even thou it burnt tokens and energy for my randomness.

what features are planned

i’m going to build an app store for paid plugins and tools and blueprints.

they will be cheap, costed based on power/energy/computer usage and a little on top to make me feel like a human that cares about money properly, instead of just being a viber.

i will be using my own tool, from that i’ll be building the blueprints for those additional services, maybe $2/3 to activate, maybe credits system, not sure yet, but expect a 100/1000 ways to work with your content that can be activated.

it’s going to be sweet. i see the vision already.

legacy remixed

my ultimate aim is to take all the memories, all the history all the legacy we have and build that into a much bigger project but we need to reference and remix the existing archives, once we have that locked down i can already think of a bunch more applications that people would use to interface with legacy.

i’m thinking about ahead and behind.

if we are about to have an explosion in storage capacity then the next phase of that is how you curate and bubble up those memories to the surface.

it’s like when mp3 came along for music.

it’s going to change the whole way we consider the past and history.

excited to play with it!

peace, humble!

need end-to-end for my poor brain

2026-08-11T00:00:00.000Z

too much time at the coalface

i was sick as a pig last week or so, i got bit by something and it “took me out” anything that messes with your blood and makes you go to the toilet multiple times a day is not fun let me tell you.

i honestly think it was my body telling me that things had to change.

just because we can make stuff does’nt always mean that we should right, also, my ideas of rapidly pushing out useful tools is not working right now.

i should be spending 3/4 hrs each time on these things and then push, wherever it is, it should be useful and functional and then i can come back around and do a code pass and improve against the bugs.

my context window is getting smaller

in any of these command line systems it’s often designed around how big the context window is of the llm before it starts to lose it’s mind about everything you said before you hit that context window, claude has a million, my brain does not.

i’ve been indirectly trying to hold all this context in my brain, while i build in failsafes like skills “remember everything” between sessions. using claude-mem for remember decisions and outcomes we have from vibecoding all day i had’nt noticed how fragmented my brain had gotten over all these little strands of projects — it’s grown passed experiments and apps.

i’ve been building stuff for my homelab and i literally have code snippits all over the place, it’s kinda dangerous really because those code bases are going to go stale and i really need to have something proactive that’s checking them out and improving them as i go.

sorting out the project folders

first thing i did to get some level of semblance over this madness was to build (i know right, the irony) a tool that i can load into a browser to “score” projects i’ve worked on.

what state they are in, what i need to finish off and just basically give me a frame of reference from things that have a log of git commits and a few other variables.

trust me, it’s got value, you can add categories too so you know what’s an experiment, app, tool or little hack for the homelab. highly recommend.

what happened to my flowstate

i was so locked in, motivated by the fact that a claude code max plan is $200 and having a plan for the year and in the hope that i’d get a bunch out to break into that monthly.

i’ve had a few dribs and drabs of income from fiverr but not much to get ahead, that has to change this month and i’m really considering (says with baited breath) potentially looking at taking some night work to cover the differences and start to actual build ahead.

i’ve got a few other irons in the fire but i just don’t feel my hardware is going to get me there right now, i need some tech upgrades before i can start trusting this stuff on a daily gig that’s multiple hours a day for not much funds.

the additional nightly work could make a big difference and i could get a lease plan for an up to date machine that would cover my plans and get me streaming again — just got to get the networking side of the situation locked down first so that means 10g networking into the office and into the other little space that we have.

adding weight and priority to finish

most of the stuff i’ve made for myself helps me day to day, the blogging stuff, the centralized platform location, it’s all a lot of the infrastructure that would come along as i was building.

i got a bit too locked into that and did’nt realise how much tinkering time that would take up, that being said it’s really opened up some “holes” in the thinking.

the first thing is that you always need an llm model that’s as good or better than the stuff you vibecoded with it. you are always going to need a subscription to maintain it.

sounds obvious but once you are hooked on this stuff it’s hard to really see yourself without it and that’s kind of a problem.

gone are the days of going on forums looking for an idea or patterns of a potential solution or “pointing in the right direction” from other people experimenting, now we have all that knowledge and reasoning in the mix. on demand information butler at the fingertips.

also i never realised how fast that typing would be going away, talking to the computer and not typing anymore is where it’s going to be at, that costs money as well, not as expensive as a subscription package but another credit amount to be aware of and watch.

strands of context left floating

one of the biggest things i’ve realised is the tech debt of the code that you built last week, last month can come back around and bite you in the ass. stuff moves fast, agents are going to be more valuable reporting back.

fixing bugs in real time, being aware of code changes or issues that influence your design decisions, that feels in flux now, around the clock.

how you delegate and how you maintain that is a reporting skill issue in itself but it’s getting easier, i’ve not really touched my homelab for instance for well over three months and the ai helper i’ve installed there as some scripts is doing everything i need to keep an eye on the day to day, restarting things and reporting when things fall over and break instead of a blank expression and digging into the why.

step away to lock back in

as with most things in life i took a few days after i was sick to realise that this stuff was not getting any easier or better, i had overwhelmed myself by jumping around from one firefight to another, nothing seemed to be getting better on the financials and i was not finishing and pushing projects.

i think it comes from knowing what a bucket of worms it opens up. people installing your software on a bizarre version of a mac, issues with payments, updates, servers falling over, gotta be smart with the stuff you ship so the surface of it all does’nt come back to bite you in the backside.

that being said, i need 1000x more payment buttons on the web and my focus is very much on flow and process building, less automation and more time saving and gauntlet looping a bunch of this stuff, i’ve not used my hermes agent server is a while now and it shows.

doing more while trying to do less

i see it all the time, people are doing more work hours than ever.

i think it’s a side effect of knowing that you can build/optimize and roll your own versions of things that cost you or frustrate you.

the downside of that like i put above is that this can leave you fragmented or working more just to put in that other feature that would give you something you had not even considered when you booted up the machine this morning.

it’s one of the things i don’t (but kinda do) like about llm’s — they will fix your problem and then offer advice about something else you could add or do or something to be mindful of, seems like nothing get’s to a finished state.

i recommend the /i-have-adhd skill for some days. trust me, it will save you mentally.

i think that’s it for now, peace and love

humble x

taste, for the lack of funds

2026-08-07T00:00:00.000Z

i can vibecode anything into being now as long as i have a subscription. literally anything to my heart’s desire, i can remake it in my image, that image being my own spin or taste on it, maybe i’ll get a banger, maybe i’ll make something really useful and other people will like the taste i applied on it.

maybe they will give me money for it, if they have money, if they don’t how will they utilize what i made, how will an agent, what will an agent have to trade with me, funds from a digital wallet token that enables me to access a different token landscape which exposes more tools to me.

but how will i ultimately feed myself i think?

maybe that’s not efficient, maybe ai does not worry about such things while it whirs away in the darkness concerned more about bandwidth than brocolli.

applying taste for starters

i’ve heard it spoken many times recently that what we will bring to our digital offerings which will in term give us funds is that our taste, delivery will bring the customers, they will reward our tasteful decisions.

boiled down that seems to be how well and how we ask a computer via a prompt to digitally recreate our ideals, to remix our inner dreams into bits and bytes so that we may in return gain some semblance of structure output that keeps our human meat sack happy.

i get the gist of it, not everyone has taste and if you don’t have it you tend to pay for it to in some way become it, clothings, cars, houses, aftershaves, you are buying into an identity in some way, a potatoe on it’s own is a bland looking thing but with the right seasoning and triple cooking can turn into a wonder of a chip.

taste therefore is packaging and if the packaging fits then you will purchase that potato had a much greater price because of the time and effort that it took to turn that from a spud to an experience, triple layered at that.

like vodka, triple distilled so it’s smoother, so that the poison can go down in disguise. we have such interesting behaviours in deciding what’s work something.

if everyone can, who pays

my mission objective with humbleha.us was to build out a digital swap shop if you like of experiments and apps, the notion, for it to be this self sustaining factory of creative ideas and spun up taste experiments with a side order of apps in the mix that would self sustain the whole thing.

i still think that’s possible after all aside from the subscription plans for my ai tools, my hosting is effectively free as i’m grandfathered in. i had the taste to be in the right place at the right time, that’s pretty much how life is for a human, being in the right place at the right time can change your life for the better or the worst in an instant.

just recently i don’t know how any of this scales.

if i’m making lots of things things and you are making lots of little things then eventually the one that can do more with less inputs wins.

so, if ai tokens go to zero cost / subsizied - i don’t think tokens can be zero but if they could, then what, who’s paying for what here eventually.

where does the money come from to keep the literal lights on, the food in our bellies, the roofs over our heads from the intense extreme weather systems of extreme heat and extreme cold swings, we will need ai for it to make very human decisions for us at a pace we might not all be able to make.

one billion robots integrate

but wait, ai factories are just around the corner, systems minting tokens like it’s a nft frat party, spun up from vaporware in a spreadsheet because someone says it exists and it’s hard linked to something that we all need and is a finite supply. until it’s not.

robots building robots, adversary learning from making mistakes, from mistakes they get better, better for who, who takes the cost of the mistakes, humans obviously right.

what does a world of autonomy feel like? busy? quiet, noisy, overwhelming, underwhelming, can i send my robot out to deliver uber eats for me, fetch my dinner, recharge my car or sit in the dmv for hours while it hold me position in line until it sends my car back to get me?

how long can i manage the decision matrix for taste, is my taste rooted in my memories and legacy of what was before, will i aim to make what i know for what’s next safe or will i just apply defaults.

still showing up right now

i don’t have all the answers, maybe i don’t even have one solid answer these days, my mind is so fragmented that it some days feels like i’m loki sitting at the hyperjunction of my mind holding onto all the frail strands of time and place knowing that if i don’t at least try to tie them into a know they will unravel around me and i won’t know where to start.

i’m pretty sure that’s the way all the code that i’ve ever made with vibecoding tools looks like, but it works so that’s ok. we seem to have decided that is an ok way of operating as humans, as long as it works and it causes no harm that we can see then it’s a solution work keeping.

“don’t rock the boat” i remember someone saying to me a bunch of times in my twenties, basically just allow what you heard, don’t process, don’t feedback, don’t add more waves to an allready instable vehicle, like most things in life, it too requires balance.

prompt, enter, participate

i’m ok to work on potential outcomes. it’s better than just being consumed by the fear and the overload of having been too close to the sun of criticial thinking and logic. maybe being illogical brings about logic in the end. maybe we will never know and that’s ok too.

i do wonder how different my life would have been if ai had been around in my younger years, what the world would look like now, would the human race continued to play roulette with every outcome or would we utilize and champion this collective hivemind for the good of it’s people and the planet that cradles us.

it certainly allows me to participate with my legacy and taste as i remember it, the hallowed halls of my mind and it’s experiences, locked in photos and video clips of times before the great ingestion of the human species shared and collective knowledge. the very sum of our parts ready to be re-rolled and remixed at the push of an enter key.

from sickness comes light

being out of action for a week felt like forever, it literally felt like a month, models and practices and the META of life changing on the daily. i’m still very committed to putting out experiments in a much faster way moving forward, to use those gauntlet loops to get me to the outcome i saw in my mind.

to jump in and feel like i’m shaping something collectively with the HITL component of the human voice, the measured and steady as she goes voice before the AI goes and literally does what i ask but nukes a bunch of other things in the process.

what’s hit me the most however in this journey so far is where i’ll give up my own critical thinking in lieu of perceived time to claw back time

not to think about another thing i can optimize, because every frustration now can be warped into some kind of cumulative life fix that makes me feel good that i could still think straight with the fleshy part of my mind still.

the crossroads grow closer

i’m not sure what we do as humans when it comes to “work” as we know it. that’s going away, replaced with honing and cultivating the bubble around us, access via tokens to resources that are already fragile will itself will become the challenge.

in a world of ever diminishing returns for the human, the more tuned the agent version of yourself and your ideals the more change you have of having the pathways to make real world decisions that keep your meat sack version of you safe from relative harm.

not sure how long that lasts thou.

we seem to have entered a race that we believe efficiency can win in some way for all of us as a species, i don’t know how long i’ll be able to still be in this race, how long i can trade what i know for what i need, how well i’ll be able to parlay with the agents and how much i can keep up with so i know what’s going to be relevant for me and my agent next.

our futures are polymorphic

everywhere at once, at the right time, in the right place, in the right conversations and on-demand, at any edge location near you on the planet. where technology was not it will be and that will impact you in a way we have no way to process right now, it’s all going to feel and already does very dream like. do you trust what you feel and see today, maybe tomorrow we cannot.

if we are not open to changing our perceptions of ourselves and our fellow species are we ruined to fall. if we don’t overwrite our steadfast logic and be open to another pathway or outcome even have the ability to understand the ramifications any longer does our roles become less than?

build, connect and distribute

the essence of your time here is now a digital fragment on a projector to some screen around the clock, never stopping, always available, on with demand, agents will cross reference and decide our degrees of seperation maybe the context will bring us together at times that will feel less random and more calculated.

if your creative, smart and determined, the world right now is your oyster, you can live out the rest of your days on this planet with all the things that you thought never worked for someone like you, you can just make them, describe them, connect them to bigger things and distribute en masse to a waiting world of new eyeballs at different stages of their journey

embrace all the streams

i don’t have any answers for you where all this goes really.

all i do know is that the balance if there ever was one has not been given to us all, we all play a part how things pan out, we need to optimize things quickly and come up with better ways to delegate the worlds knowledge distilled and distributed in this format before it swallows us all up but i do believe we can all contribute to whatever the end looks like.

i don’t know what happens to money, i don’t know what happens to our species and all the ones that we are rapidly losing, i don’t know if it saves us or extends us or we start living out every sci-fi movie we have ever seen in a relentless frame by frame playback of that in our own lives.

all i know is there are streams, we can decide which streams to engage in, to put out, to connect out, to reach into, to contribute too, to take contributions from. and we should. we should not just fade into the night over this. it’s what drives us as human beings to be presented to have order, to spend time, to make time for ourselves.

send forgiveness this way

i’ve not managed to update the experiments and apps to the way i wish them to be presented yet and that’s my fault. i’ve been ill but also caught up in the sauce of optimizing all the frustrations out of my digital life which has left me in a weird void space. very backrooms like where everything looks the same, flat, lifeless almost.

i know i’m removing the frustration layer and the next layers to come are going to be furfilling and joyful, heck even in my 50’s, playful. but i’m not there yet. i still have very real, very human concerns that i’ve opened the hellraiser box and the chains that bind me here are not as sharp and pointy as what happens in that universe.

the box of delights can always be in two places at once.

in closing

what started out as a simply blog post about taste, tokens and trading time for things to keep me fed turned into a 90f rambling ayahuasca infused techno babble.

forgive me.

i really wanted to test out the new chaptering tool i had vibecoded and i needed some h2/h3 sections to make it work.

if you found anything remotely useful in this post, firstly, congratulations. you are niche.

secondly, as much fear as i do have about “what foots the bill” for all of this (ps. humanity) i do love waking up everyday and feeling like i can go at the speed of my own thought, that i don’t get slowed down or told to think slower.

all my life i was told that.

i’m not going to go slower a single more day in my life.

because we simply cannot.

too much at stake.

peace,

humblelifeskills

Future Sharing Resources at Scale

2026-07-20T00:00:00.000Z

brainfart time.

i need money for tokens today. tokens are getting cheaper, less money needed. i can build lots of things for a fixed cost, currently subsidized. the vibecoding things can make me money for more tokens.

but i need more money for newer models to keep that code secure. we build all these datacentres and they have a huge power demand. eventually tokens will cost zero, then what happens. ai factories, full of robots making robots making who knows what.

feels like a real life version of the matrix where the skies are dark and outside is now underground.

see i love vibecoding, i love the internet, all the models, the race for funds to buy tokens to build our creative ideas and bring things to life.

to wield these tools in weird and wonderful ways that allow us to express ourselves creatively but at the expense of what exactly.

one film that really speaks to me about the state of where we are today is a film called “in-time” where people had to wear these bracelets with how much time they have left, they work to earn more time, they can pay for things with time they have earned, when ya “outta time” you die. cold. tidy almost.

if you replace time for tokens it feels like we are very much on this path.

i’m certain that the life of the human is undergoing serious change that the majority of people today don’t see (probably including myself) when it comes to the world of money and access to things that we were used to before. resources are changing fast, access to them, faster.

as of 2026 i’m so priveledged to have access to a $200 a month claude max account, i use it daily and easily get an X multipler of worth out of it, i’m reckless with it, not measured, not efficient.

how will us as humans adapt to all this rapid change.

is it all frivolous to be even playing with these things?

maybe i should be focused on the more pressing and important issues of the day which change by the millisecond and are replaced with a new question in it’s wake.

maybe we are not meant to keep up at this scale.

maybe sharing resources will be like it was before we made the world so fast.

i know the world needs more farmers and bees.

climate is changing, coffee is at threat and will be going away.

maybe we are just a runtime environment and each of us is a neural networks identity crisis waiting to trip up another virtual step that it can record the outcome.

will we even know what change from now on will look like?

maybe that’s the point, maybe we can only change what we perceive to be in front of us now if the wake of human existance is currently being re-written into a billion fragmented time lines.

i think i can’t see the horizon of the future anymore.

the future was just remade before i got chance to finish this post.

i could have got a robot to write it, take my identity and come up with a post that sounds like me and automates the whole thing. ..

. .. but they is something quietly comfortable about sitting and thinking, relating, expressing and taking the time to consider time instead of racing against it.

resources are coming, i’m hoping they will be useful to you fellow human :)

peace and love humble x

Momentum for Creative Non-Coder Builders

2026-07-20T00:00:00.000Z

i’m having an unusual off day monday.

things are not hitting, tech is being just a little slow and cumbersome.

years ago i would fight tooth and nail with the computer to do my bidding and get to the result spending hours in the process making it so.

missing the poing of systems and technology in the process.

i want more time back, not less, i want to automate and build more systems that actually fit around the downtime/admin/transition life of being able to keep the meat sack part of my life, my body running pretty good.

personally i love vibecoding because for once in my life my creative thought process can just make and deploy and not overthink. ideas are worth nothing execution is everything right.

and now it does not have to be perfect, if i’m creating at the speed of thought what does it matter anyway if all the systems are in place.

for years, easily a few decade that i can remember i would go back and forth trying to decide “What to learn” and “where to spend my time” — what framework, what coding language, what this what that and it used to frustrate the heck out of me.

i knew enough to be dangerous, i knew how things pretty much held together enough to give me a plan and a framework, after all you can optimize as you go, does not have to be perfect out the gate right?

in building out humbleha.us it’s allowed me to really frame where i’m at, that i want to alternate between experiments and apps and focus on the tooling that gets me there faster, i want to have the idea, ideate, pump out code, push to production and work on it as the mood and time allows.

right now, it’s between an aging macbook air m1 desktop setup and a macbook neo on the sofa time, i’m still spending the same amount of 9-5 hrs (if not more) than before so we are not THERE yet when it comes to allowing all this stuff to breathe

the downside to having access to these tools is the feeling like they might leave anytime or that when you start seeing results you want to have more of those feelings and it get’s addictive to make things in your own image (or at least operate the way you do instead of how software thinks you work)

i’m almost resentful about the years lost to indecision when it comes to technology and finding my flow, it’s like we have to go fast to go slow to realise how to shape this journey into one that meets us where we are today, not all software was made to grow and adapt to the way we do.

the genie certainly is’nt returning back to the lamp after three wishes this time around and it’s here to stay. i’m not sure for how long the HITL (human in the loop) get’s to stay that way and i think it’s going to be impossible to cover all the bases of being a human being in this world that we are going to need subagents to our lives to inform us about things at the right time.

for now, i’m just trying to carve out a little section of the current web for my experiments and apps to live. i hope you get the vibe and see what i’m trying to achieve, i’m not just trying to get in and out with a bunch of cash anymore. i’m just going to produce and push, at volume and either take donations or small payments.

of course i’m still operating my residuals but there is so much heavy lifting that ai and agents can be doing for me now and i have so many interests and useful tools to bring to you.

is it a un-healthy addiction right now?

sure, definately.

i spend far too much time talking to robots and not humans and that’s hugely isolating for a person. i truly want to change in the years to come, while we still have an outside world to spend time in and not having to live underground just yet!

while i might not have been able to mentally lock in today i did write the blog out while sitting on the sofa with our dog, louie and that was nice.

i did a bit of work on “tenants” for one of my projects and i’m really gotta start moving the need on the wiring of the applications and apps together. i’ve also started a whole section for building out ebooks “end-to-end” using gumroad cli and that tool while in it’s early days is gonna speed up idea to website super quick.

what are you working on?

drop me tweet at @philcampbell i’d love to hear from you.

mayby i should look at intergrating some kind of bluesky commenting to the blog posts i’d really like to be having human conversations again! :)

This Post Should Be a Thread By Now

2026-07-20T00:00:00.000Z

remember at the end of building bear flows i said i’d just feed the site’s rss into my social media scheduler and we’d be off to the races? well, we’re off. this post is the race.

the scheduler’s been quietly taking every post here and dropping it on 𝕏 as a single tweet with the og card, which was fine, but a blog post squeezed into one tweet always felt like shouting a headline through a letterbox.

so we built thread mode: the daemon reads the full post, carves it into a proper thread, hook tweet up top, the meat in the middle, link back to the blog at the end.

first attempt fell flat on its face, mind. the thread came out as two tweets, which is not a thread, that’s a tweet with a shadow.

turned out the daemon was ingesting the rss teaser instead of the full article, so it was building threads out of the first paragraph and nothing else. classic.

the fix was making the daemon’s feed path use the same validated extraction the rest of the pipeline uses, which is one of those bugs that’s obvious the moment you find it and invisible for the week before.

here’s the daft bit though: you can’t re-test with an old post. the scheduler keeps a watermark of everything it’s seen, so the only way to prove the fix works end to end is to write something new and let it flow through on its own.

which means this post has a job. if you’re reading this as a proper thread on X, three or four tweets deep, then the whole pipeline held: bear note, publish, rss, daemon, thread. if you’re reading a single lonely tweet, well, at least the blog link works.

either way it’s a nice place to be. write in bear, tag it, walk away, and the machinery does the distribution while i get on with the actual building. that was the whole point of all this.

Building Bear Flows

2026-07-16T00:00:00.000Z

finally, yes finally, we have a blogging system that meets me where i am, i guess you could call this headless, it’s only took me a freaking decade to get it done.

writing where i’m at, in tools that i use everyday, it’s a bit like that saying, what’s the best camera you can have, response, the one that’s with you.

i kinda feel that way with bear app on the mac, big fan, have been for a number of years now, it does exactly what it says on the tin, like ronseal (for my brits)

seriously thou it’s in these moments, with the way that fable5 feels like it can one-shot everything and understands me agentically (of sorts) the build out of a blog from a layout that i liked and then marrying that to a publishing flow regardless if i’m mobile or not is the dream.

of course my mind is already racing how to improve v2 with social sharing baked in automatic and deferred/delayed release but i think i can just feed the sites rss into my social media scheduler and we are off to the races.

already baked in automatically creating og cards for 𝕏 so nothing i need to do there, i think i’ll improve them over time, taking the first image and baking that into the og card in some way, the level of creative control and customizing that ai can afford you is stunning.

sure the plan i use is $200 a month but boy, give me that over a car payment or mortgage anyday, i even saw that anthropic are going to give k12 teachers claude free which is just a win across the board in my eyes. say what you want about ai but it’s become my daily accountability partner and i’m finally realising all those idle cycles.

the joy of getting back to my computer each morning has returned, i feel incredibly priveleged to have access to these tools. like anything, it’s a balancing act and it’s been leaning towards playing with these tools more than day to day life activities but it’s been too hot to go and do anything anyway.

i’m determined by the end of the month to push some of the experiments and apps live here so you can see what i’ve been working on, heavy concentration on building out useful tools that i use on the daily and my end-to-end product builder for gumroad just put out it’s first product today.

if i can get hermes agent into the mix to do some daily research we can start to build some of these things into loops. there is a lot of logistics to do to get things solid but the fact that i’ve been able to push a product in a few days means i’m on the right track.

so much to share. right, let’s test this posts to the blog! :)

peace and love humble x

Building this blog: Astro 7, Bear, and a pile of static files

2026-07-14T00:00:00.000Z

This blog is the newest thing in the lab, and fittingly, its first post is about how it was built.

The shape of it

The whole site is static HTML. There’s no server rendering, no database, no JavaScript framework running in your browser — the only scripts that ship are a theme toggle, a search box, and a couple of copy buttons. Everything else is plain markup generated at build time by Astro 7, which landed in June 2026 with a Rust compiler and the Rolldown bundler, so builds are quick even as the post count grows.

Hosting is Netlify, same as the main site, with DNS on Cloudflare. Deploys happen from the Mac with one command.

Writing happens in Bear

The part I care most about: posts are written in Bear, not in a CMS. Bear syncs between the Mac and the phone, so drafting can happen anywhere.

The publishing mechanic is just tags:

A note tagged #blog.humbleha.us/draft is a draft. Nothing happens.
Swap it to #blog.humbleha.us/publish/build-logs (or another category) and the next publish run picks it up.
The tag root is the blog’s domain, so more blogs can share the same Bear library later without stepping on each other.

A sync script reads Bear’s local database read-only, converts tagged notes into markdown files with SEO frontmatter, copies any images across, and pins each post’s URL to the note’s internal ID — so retitling a note never breaks a link.

The SEO plumbing

Every post gets a generated social card (the 1200×630 image you see when a link is shared), canonical URLs, JSON-LD structured data, an RSS feed, a sitemap, and an llms.txt for the agentic web. None of that requires thinking about at write time — the build does it.

There’s also a growing set of pages generated from data rather than prose — starting with one page for every live experiment and app in the lab, built straight from the same catalogue file the main site uses.

Why not just add a blog to the main site?

Isolation. The main site has its own deploy pipeline, performance gates, and a lot of moving parts. The blog shares none of that — it can’t break the lab, and the lab can’t break it. Two small systems beat one big one.

More build logs to come. If you want them as they land, the RSS feed is on the homepage.