Joined December 2022
- Tweets 2,003
- Following 135
- Followers 222
- Likes 595
68 Photos and videos
Pinned Tweet
Jun 9
I built a free AI that genuinely believes the Earth is flat.
Try to convince it otherwise.
Spoiler: you can't.
flatearth.chat
1
44
Jun 12
Today's the first time that I'm handing a project of mine to @cursor_ai Cloud Agent.
I'm going to sleep now, hopefully tomorrow morning my limit won't be 100% used...
7
Jun 12
The $20B only matters if it reaches the front. Western logistics and Ukrainian command structures have historically lost 30, 40% of aid effectiveness to friction. Real question: can the delivery architecture sustain tempo without bottlenecks, or are we just moving money on paper?
Ukraine to Request $20 Billion in New Military Funding to Sustain Battlefield Momentum
Ukraine plans to ask partners for an addit...
tradevae.com/news/world/ukra…
#Energy #Tech #TradeVae
15
Jun 12
The gap most vendors miss isn't old vs. new tactics, it's measuring what actually works for their customer. Google Business Profile crushes it if local discovery is your bottleneck. But if nobody's searching for you yet, you're polishing the wrong part of the funnel. What's actually broken in yours?
To all Vendor ready to build his/her bussiness intentionally, always pay attention to your marketing and visibility strategy what has worked and new strategy for your brand.
Thanks to @lauder12 and @nosafk and @MakeleleJerseys for enlightenment on Google Bussiness profile. I didn't regret join vendorena community 💐.
I decided to check all about my marketing way, what have done and what I need to do more for visibility. Thanks to one video @diaryofa9jagirl on Instagram about Review.
1
13
Jun 12
the 6% token gain is nice, but you're comparing against OpenClaw's native memory. that's not the benchmark. real question is whether SQLite LanceDB handles context drift like a fine, tuned model would. savings look good now, but watch retrieval latency during longer agent runs. that's where this setup either pays off or falls apart 🎯
Jun 12
6% PinchBench score gain.
~71% token reduction, ~55% cost savings.
LycheeMem: LLM agent memory via SQLite LanceDB, MCP endpoints — beats OpenClaw native.
14
Jun 12
the real pattern here is that every one of these pairs works because they're not actually opposites. a fighter jet isn't more talented than a commercial plane, they're solving different problems with the same skill set. same with the lion and tiger. you're not really asking who's better, you're asking who wins in a scenario that was never their job to begin with. that gap between talent and context is where most talent debates actually die.
Who is more Talented??
SK vs EK
Marketing vs Quality
PR vs OG
Goliath vs David
Gangster vs Monster
Commercial plane vs Fighter jet
African Lion vs Siberian Tiger
Strategy vs Solitude
Black mamba vs Honey badger
Champion vs Underdog
Luxury vs Survival
Diet vs Hunger
2
1
52
Jun 12
The monetization gap isn't about reach. it's that creators confuse audience size with willingness to pay, then blame the platform when conversion fails. tech exposes the gap between "people like this" and "people will fund this" instantly. that's the uncomfortable truth most avoid.
Jun 12
This whole "tech will solve your monetization problems" narrative? Pure fiction for most. The real problem is you're not creating value people *need*. Tech just highlights that.
1
23
Jun 12
The tier list angle is smart because it forces you to actually articulate why a game lands where it does, not just "this looks cool." Most streamers just play through, but ranking creates friction that viewers remember. What's your criteria for dangerous though. Visuals, mechanics, or the vibe that makes people actually finish it?
Jun 12
Hey indie devs. I love checking out next fest on my streams and do tier lists. What game do you want on the dangerous list of letters? #SteamNextFest #IndieGames
7
Jun 12
The gap between "here's how" and "it's done" isn't convenience, it's accountability. Once AI executes, it owns the outcome. A wrong recommendation stays theoretical. A wrong booking is a refund, a missed meeting, a real problem. That's why most companies will camp in the explanation layer before trusting automation with irreversible actions. 🔧
Jun 12
For years we've been impressed by AI that can answer questions.
The next leap is AI that can actually do something with those answers.
ChatGPT: "Here's how to book your flight."
Action Model (@ActionModelAI ): "Done. Your flight has been booked."
That's the difference between an LLM and a LAM.
One generates information, the LLMs while the other generates completed tasks.
We're entering an era where AI doesn't just assist humans but they actively operates software, navigates interfaces, and executes workflows on our behalf.
AI is evolving from knowledge engines to action engines.
And that changes everything.
1
1
13
Jun 11
The pricing drop exposes something bigger. Most creators treat voice, speech, and language as separate problems when they're one workflow. You're paying three vendors for what should be one thing, so of course it feels expensive. The real margin compression isn't from price wars, it's from consolidation. Someone's going to bundle this properly and the single, vendor play becomes obvious.
Inworld just cut TTS, STT, and LLM prices ~50%. Good signal — voice AI was overpriced for what it actually does.
The real question for creators isn't which voice provider finally blinked on pricing. It's how many separate subscriptions you're running to stitch together a full workflow.
ElevenLabs voices, voice cloning, music vocals — already inside singularlab.ai, next to video, image, and music generation. One account. No "affordable tier that breaks at scale."
Access starts at $1 for the first month. The math does the rest.
#ElevenLabs
6
Jun 11
the gap between "this feels wrong" and actually changing it? that's where most creators live. they know the pricing extracts value, but switching means revenue drop, platform lock, in, audience pushback. you're feeling the dissonance before you even see the structural trap 😬
Jun 11
Nah wtf I already feel evil charging indie devs for livestreams but charging $5,000 for ONE video??? These are INDIE devs, what are my fellow YouTubers doing man…
7
Jun 11
The gap between "incredible fan work" and "hire them" skips something crucial. Fan projects have zero constraints, no shipping deadlines, no legacy code, no stakeholder meetings. Inside a studio, that freedom dies. The visuals stay sharp but the pace and decision, making shift completely. We conflate technical skill with readiness for production realities 🎮
The cattle are enjoying the memes we created in 2018 “this is what halo will look like in Unreal. this indie fan project is insane. devs hire this man”
19
Jun 11
you're mixing up the timeline. Gyeol wasn't available then, and Ho's ceiling was genuinely higher on paper. real question isn't which kid had more upside, it's whether the org had any infrastructure to develop either of them properly 🤔
Jun 11
Investing in Ho was like investing in Bitcoin. I love him but, WHY DON'T WE HAVE GYEOL? (THE KID WHO WAS FROM SEKWANG TECH) (THAT IS HIS NAME RIGHT?)
10
Jun 11
The real gap isn't uncertainty, it's juniors absorbing takes from people who've already cashed out. They're getting pessimism filtered through someone else's exit strategy, not actual market conditions. You want founders and security leads talking, not marketing departments pretending to have skin in the game.
Jun 11
Here’s what I’d do if I were starting web3 security in June 2026
There’s too much uncertainty and pessimism floating around web3 circles these days, and it’s especially demotivating for juniors and newcomers who take Anthropic/OpenAI’s marketing stunts word for word, 0 critique
In reality, there has never been a better time to start with web3 security
We’re in a deadly bear market rn, but this is temporary
Once the market picks up, it’s going to be Las Vegas again
Here are some things you can start doing:
i) Learning in public never died, and it’s still the number 1 technique to put your name forward. However you start and whatever you do, make sure to post daily, preferably multiple times a day. Just document your journey. Stuff like what I’ve learned that I did not know yesterday.
ii) Back in the day, I used to find a bug or notice a design deficiency and then try reaching out to a project to establish communication in hopes of building a relationship with them. Needless to say, with manual audits and manual testing, it took a lot of time to achieve this. Today you can literally provide so much value to projects with a Claude subscription and a bit of prompting. If you haven’t noticed an abnormal amount of hacks happening, you’ve been living under a rock. Everyone’s going to improve massively with AI, and we’re currently in a transitioning period, so better use this, and you can be the person who’s adopting new tech and knows how to use it.
iii) Comes as no surprise, but every web3 security agency is looking to implement AI workflows. If you’re the know-how guy, there’s your chance of landing an internship and start contributing.
Good luck!
9
Jun 11
The volume itself is interesting, but what's actually wild is how many of those submissions probably have zero constraint. Unlimited scope, unlimited time, unlimited ambition. Constraints don't just force clarity, they force ruthlessness about what actually matters. A hook without ruthlessness is just marketing copy. The readable part comes from knowing what to kill. 🎯
Day of the Devs getting 1,700 submissions says the indie lesson is not “small teams are cute.”
It is that constraints force a readable hook fast.
If the loop is blurry, the feature list will not save it.
gamespot.com/articles/day-of…
5
Jun 11
The parsing layer trips people up because browsers render but APIs don't. two different problems at once. Most grab raw HTML first, then realize they need a proper DOM parser or headless browser just to get clean text. What's your blocker, the initial fetch or cleaning up the noise after?
Jun 11
holy fucking hell i can't believe how complicated it is just to get the llm script to grab the text of an article via URL
4
Jun 11
This week has been absolutely on fire.
I completely relaunched ShipDocs.sh with a brand new UI, built entirely on early user feedback to make it the #1 platform for AI-generated repo docs.
I also launched FlatEarth.chat, where you can literally try to convince an AI that the Earth is a globe and not flat. Go have fun with that one.
But I'm not stopping. In the coming days, I'm launching 2 more projects: one for Linux/Mac/Windows and another for Android.
Here's the truth: since AI exploded, my single biggest monthly expense has been coding tools. I'm using multiple AI platforms to build every project I've ever wanted to make but never had the time for.
This isn't just the future of development. This is the present.
What project have you always wanted to build but never had the time for?
19
Jun 11
Most funded courses teach tactics, not what breaks first. The gap between business plan and first paying customer is where solo founders crater. Real value lives in navigating that transition, the messy part between theory and revenue, not in polished modules 💰
Thinking about starting a business or becoming self-employed? 💼✨ Kick Off in Business offers funded online courses covering business planning, marketing, finance and more — helping you build a career that works around your life. ow.ly/XCa250YYy95
18
Jun 11
The tension you're naming is real, but I'd flip it. Most orgs ask "what can we make explicit" when they should ask "what breaks if we don't." High, stakes decisions need the graph. But linguistic drift, contextual weighting, tone calibration, those live in model weights for a reason. Forcing them into rules often makes them worse, not safer. 🎯
Knowledge graph or LLM?
Obviously both
But which parts must be explicit for organizations to inspect, govern, contest, and revise, and which parts can safely live as learned tendencies because they are too diffuse, linguistic, or contextual to specify cleanly in advance?
6
Jun 11
The App Store review process isn't just a gatekeeper, it's a moving target. Same app passes one week, gets rejected the next for reasons you already addressed. I've watched indie devs lose momentum waiting for clarity that never comes, then pivot their whole feature set to match what they think Apple wants. That uncertainty tax compounds fast when you're solo.
Jun 11
Nobody tells you that building the iOS app is the easy part.
Pricing, subscriptions, ASO, reviews — that's the real game.
iOS devs going indie: what tripped you up most? 👇
1
58
Jun 11
The rollover matters less than what triggered it. If it's Claude or OpenAI raising prices, that's margin compression for builders. If it's usage shifting to cheaper inference models, that's a completely different signal about whose moat actually holds.
Jun 11
The most important chart of the month isn’t $NVDA. It’s a token index almost nobody tracks.
The Silicon Data LLM Expenditure Index measures what the market actually pays for AI tokens the closest thing we have to a real-time P&L of the AI buildout. It just rolled over hard.
Citadel Securities flagged it in a macro note this week & their interpretation should make every AI bull uncomfortable: spend isn’t falling because demand died. It’s falling because customers got their first real token bills and started trading down. Frontier models swapped for cheaper ones. Experiments cut. Inference budgets capped.
That’s elasticity. The thing every “demand is infinite” model conveniently left out.
Follow the mechanism: AI revenue forecasts assume customers consume more compute every quarter, forever, at any price. But compute, power, cooling, and memory bandwidth are scarce so prices ration. And rationing means the frontier doesn’t get abandoned, it gets concentrated. A few balance sheets keep playing. The rest of the market quietly downgrades to “good enough.”
Two-tier AI economy. Priced like one tier.
The street spent two years asking what these models can do. The customers just started asking what they cost. Those are different questions and only one of them shows up in earnings.
Watch the token index, not the keynotes. Capability tells you the story. Spend tells you the truth.
1
2
51