@lightfidreams
Account based inTurkey
About this account
- Account based in
- Turkey
- Connected via
- Turkey App Store
Account-level information from X, not a live location or the device used for a specific post.
AI n' shit. Building GrowFirst, a platform that helps you grow apps :) and a few other things.
Joined November 2021
- Tweets7.2K
- Following2.4K
- Followers2.1K
- Likes11.2K
Qwen3.8-27B at 144 tok/s on an M5 Max MacBook Pro ⚡
Meet Inco Splash: our open-source inference engine, built around the model and around Apple silicon.
Up to 3× the decode speed of Ollama, 2× oMLX, and almost 4× when an agent fans out into sub-agents.
Light retweeted
The better the harness, the better the business.
Berkeley published a study this week showing harnesses, the systems that control AI agents, set the price of an answer. The right harness cuts the cost of the same result by 71% without a loss of accuracy.
Light retweeted
All the grifters are completely wrong about Jev’s architecture so I decided I’d release an open-weight version. BUT training takes time, so while we all wait I decided I’d drop the sauce.
archerhume.com/posts/jevs-ar…
Light retweeted
TALES OF GRIMWICKE — Chapter I is 22 minutes of AI fantasy animation inspired by the many anthropomorphic animal stories I read as a kid, with a dark twist:
The model: Seedance 2.5 (@ByteDanceSeed_ ) via @Segmind_ai. 847 takes generated, mostly 30-second multi-shot prompts, over 5 hours of footage. 127 clips made the cut. About $4.5k in generation, four weeks from first take to publish.
Characters and locations came out of @midjourney. Using MJ for the location plates was a revelation: the cinematography and art direction are a level above, and it renders far better as a plate than Nano Banana or GPT Image. Nano Banana (@GoogleDeepMind run on @fal) did the heavy lifting editing those plates, building anchor frames and turning MJ characters into turnaround sheets. Then I let Seedance do as much of the framing and camera operating as possible. That gave the best results.
A lot of what I do in my Unreal Engine films came with me: a handheld, imperfect camera operator instead of smooth moves, focus hunting, and above all weather and atmospherics. The model does water, snow, smoke and rain really well, and leaning into that is what keeps it from looking like a Temu Pixar film.
The tools I built: a portal (Node on @flydotio) that holds the bible, the refs, every prompt and every take, with a Claude agent (@claudeai) that writes the prompts to the house rules. A Premiere Pro (@AdobeVideo) panel that browses the archive, drops selects on the timeline, versions the project, strips music, and sends fixes back to Seedance. Everything is done in 720p with no upscaling, because it looks terrible. Narration is my own voice through @ElevenLabs speech-to-speech. Music from @Artlist_io.
None of this would exist without @PJaccetturo and @Kavanthekid. Their narrative work, Nexus and Chronicles of Bone, proved long-form AI film was possible, and their podcasts, newsletters and X posts, updated constantly as the workflow changes under everyone's feet, showed how to actually do it.
Chapter II is currently in the works and I will release it initially on my YouTube channel, The Darkest Age (@CharlieDriscollAI). Consider subscribing there to see it as soon as it drops!
Best explanation of Jev around
this is the easiest way to understand Jev:
LLMs generate answers.
Jev makes decisions.
that sounds like a small difference, but it actually changes the entire use case.
say you give a normal LLM this:
“here’s a user, their account history, payment behavior, support chats, device data, etc.
tell me if this looks risky.”
the LLM might reason through it and return:
“yes, this looks high risk.”
maybe in JSON if you ask nicely.
with Jev, you define the possible decisions upfront:
risk:
* low
* medium
* high
manual review:
* yes
* no
and Jev returns something closer to:
risk = high (96%)
manual review = yes (91%)
that’s basically the product.
it’s not trying to be another ChatGPT.
it’s more like an AI-native if statement.
instead of:
if transaction > $10,000:
review()
you can start thinking more like:
if “does this behavior look suspicious?” > 95%:
review()
and that opens up a pretty interesting category of software.
a few assumptions I had at first that turned out to be wrong:
1. “so it’s just a classifier?”
kind of, but that undersells it.
the input can be messy real-world context, and you can ask multiple typed questions about that state at once.
fraud?
churn?
escalate?
eligible?
priority?
all from the same input.
2. “so it replaces GPT / Claude?”
not really.
I actually think the interesting architecture is:
Jev decides WHAT needs to happen
Claude / GPT reason or generate WHEN deeper intelligence is needed
normal code executes the deterministic stuff.
Jev becomes the routing layer.
3. “it can’t hallucinate?”
this one needs nuance.
if your allowed answers are:
LOW
MEDIUM
HIGH
Jev won’t suddenly invent:
“EXTREMELY HIGH 🚨”
the output structure is constrained.
but it can still be wrong.
HIGH at 92% can still be the wrong decision.
so “no hallucinations” doesn’t mean “always correct.”
4. “why not just force an LLM to return JSON?”
you can.
we already do this everywhere.
but you still deal with generation latency, schema validation, retries, weird outputs, confidence estimation and a lot of glue code.
Jev is designed around the decision itself rather than text generation.
5. “why should I care?”
because most software is ultimately a giant tree of:
if this → do that
if this → route here
if this → escalate
if this → reject
if this → ask a human
Jev is basically asking:
what if those if statements could understand messy human context?
that’s a much more interesting framing than “another AI model.”
I can see this being very useful for:
fraud / risk
support routing
moderation
PR / QA automation
lead scoring
compliance
workflow orchestration
agent routing
especially as the cheap + fast decision layer sitting in front of larger reasoning models.
early tech, obviously.
but the category itself makes a lot of sense.
Never again “structured output” from sonnet
It can’t output pure text but it can output typesafe structure, which could contain text :)
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution
Light retweeted
done, it's open source now ✅
iOS phone farm - register real iPhones, watch + control them live in the browser, schedule tiktok on a postgres-backed scheduler.
free, self-hosted, apache-2.0.
- git: github.com/Git-Agni/prod-FAR…
- DIY steps: gethandler.ai/ios-farm
OpenAI just published a guide to updating your skills and prompts for Astra. You can ask Astra to use it to review the old rules in your project.
You ask for a typo fix, but the AI first reads all your architecture, database, and deployment docs. That’s an example from the guide. You’ve changed models, but the instructions you added to keep earlier models from making mistakes may now cause unnecessary work or make Astra stop before the task is finished.
Send this prompt and the link below to Astra 👇
“Read this guide and review my project’s AGENTS.md and Skills. Identify overly broad skill triggers, documents you’re required to read for every task, redundant checks, conflicting instructions, and rules that make you stop for confirmation before the task is complete.
For each issue, quote the original instruction, explain its potential impact, and suggest the smallest change needed. List any proposals involving safety, permission boundaries, or reduced verification separately for me to decide. Show me the recommendations first. Don’t modify anything yet.”
developers.openai.com/blog/r…
Get more out of GPT-6 Astra by revisiting your skills, AGENTS.md, and task prompts.
Make skill triggers specific, load guidance when it's relevant, and define what done looks like.
developers.openai.com/blog/r…
Light retweeted
We grew from ~0 to $500M ARR, adding $250M last year alone while being EBITDA profitable.
1,300-word post on every growth tactic that worked for us:
1. What got us from 0-$400M ARR in the US works in every country.
2. Retention and Monetisation Hacks.
3. Localization > Translation.
4. Experiment with $1M internal seed checks.
5. Pivoting to only AI content production unlocked 100% ARR growth.
Pocket FM is like Netflix for audio-only dramas, with our own pool of one-person studios.
1. Acquisition Playbook for 0-$10M ARR in any country in 6 months
We run a 90 sec video trailer of an audio drama as an Ad and ask users to download the app if they're interested in the rest of the story.
Our core insight after spending >$100M on acquisition is: If the clickthrough rate (CTR) for an ad goes from 2% to 2.25%, our customer acquisition cost (CAC) decreases by ~ 30%.
We remodelled our system around this insight and built an AI-first 2.25% CTR ad manufacturing machine that works in every country.
For every new country, we take our hit shows -> use LLMs to extract and most intriguing moments -> Write a 5-min script combining all the best parts. The first 60 seconds has to have a hook every 5 seconds and needs to end with a crazy cliffhanger to force a download mid-scroll.
If a marketing video is not hitting our benchmarks (2.5% CTR and 55% 3-second through-play), a creative director gets involved to change the hook or cliffhanger to get the numbers there.
AI lets us make 1,000 Ads per show, and in total we do ~17.5k Ads per month. When a business creating scales from 1k to 10k Ads, the normal thing is for CAC to skyrocket. But with what I just shared, we 7-8x'd our User Acquisition budget without increasing our CAC materially.
It took us 8 months to figure this out, but then the timeline from 0-$10M in every country got shorter and shorter:
US revenue grew to $25M in 19 months. (US is now 78% of total)
Germany to 21M in 11 months.
France to $10M in 3 months.
2. Retention and Monetisation Hacks.
We knew we wanted to create an audio entertainment platform, but there was no standard format. For the first 2-3 years, we tried 10 different formats before landing on the winner.
After we got it right with audio drama (8-12min chapters, written for mobile fiction, serialized, episodes have strong hooks and end on cliffhangers), avg daily streaming time went from ~25mins to 150+.
Audio drama made us realize that Pocket FM was creating a whole new medium. There was no playbook for anything that we were doing. Everything had to be thought & built from scratch.
This is what we did for each major bullet:
- Monetization: Users have some free daily minutes to listen; then it's pay per episode. We also added the option for users to unlock episodes by watching ads, which is doing extremely well. Ads scaled from zero to a ~$90M run rate in 12 months.
- Engagement & production: You don't become obsessed with an app you open once a week. To have users engage daily, we make the next episode free every day. Also helps them build the habit.
- Discovery. Huge problem because people consuming Pocket FM enter the app, tap the show, and lock the screen. We fixed it by doing "playlists" of episodes. Once users's free minutes on a show are done, we ask users if they want to pay. If they say no, we play a new show, one where they haven't used their free mins. And we stitch episodes of different shows together that way, creating natural discovery.
3. Localization > translation.
78% of our revenue is still concentrated in the US. Localisation is fixing this: You might write a show for a Spanish audience where language, jokes, folklore, have a certain flavor; if you merely translate the show for, say, a Norwegian audience, that color is lost and hurts the show in the Norway. Listeners would relate more if the show was written by a Norwegian.
That's why, instead of translating, we localize shows to different regions. We use AI to take the spine of stories and adapt their whole cultural layer to the other country.
The results: a US show that was localized for a German audience had 50% higher retention than the translated version.
Localization + our user acquisition playbook led to:
- $10M+ ARR in France within 3 months
- $21M+ ARR in Germany within 11 months.
And this expands writers' addressable market.
We get messages of writers thrilled to have revenue coming from the US, India, EU, LatAm, without them doing much incremental work.
4. Experiment.
We give $1M checks to new internal initiatives, and the team has 12-18 months to prove their thesis. If they prove it, we double down.
This is how Pocket Saga came about, our AI video microdrama app. It's an AI video equivalent of Pocket FM. Same shows and structure, but as a vertical 2-min mobile video series.
We launched it two months ago and it's at ~$15M ARR.
Pocket's broader thesis is to help creators tell their stories to as many people as possible. Start with audio drama -> multiple languages or localize to diff countries -> microdrama -> more formats like movies, TV shows, and games.
Best of all is that writers get revenue streams not only from countries they wouldn't have tapped into, but also from formats they wouldn't have thought possible.
5. Pivoting to only AI content production unlocked 100% ARR growth.
In mid-2024, we pivoted to only AI content production.
Our growth took a very direct hit. Pocket was already at $200M ARR, growing 50% YoY, and we flatlined for a whole semester.
Six months later, the business exploded.
After the switch, we went from ~25k hours of content produced per year to over 2.5M hours, which are also higher quality. Because AI orchestrates the writing and more data is fed into it, more blockbusters come out.
Over 90 titles have $1M+ in lifetime earnings and 13 crossed $10M.
Quality control is done with LLM as a judge, and LLMs are quite tough.
Results are very encouraging:
- 12-month revenue retention went from 44% to 76%.
- In the past year, we added $250M in net new ARR.
__________________________________
Pocket Entertainment (Pocket FM + Pocket Saga) has become the largest AI entertainment platform.
We have the largest storytelling catalog with 770,000 titles, 550k creators, and 5.5B hours of playtime with minute by minute retention & engagement data. We are using all of this data to improve every aspect of our business.
Our Bet: In the next 3 years we will be able to produce Naruto-like series for $1000.
Netflix has to spend $17B to find 100 Blockbusters per year.
What happens when you can produce 1M high quality shows for $1B?
The streaming wars caused a $300B reallocation of market cap.
The AI entertainment wars may be a $1T+ reshuffling of market cap.
We are at a unique spot. Unlike other AI creator tools like Runway or Midjourney, we own both the supply and demand side of AI content.
We've attracted over 550,000 writers who are producing an annualized 2.5M hours of content every year. This pairs with 137B+ minutes streamed.
Because of this, growth is accelerating as we scale further.
We have multiple S-curves inflecting at the same time:
- AI produces better and more ads -> scale faster in new countries
- Writers + AI produce more shows -> more blockbusters
- Localisation -> more reach per blockbuster
- Audio to video format expansion -> larger TAM -> more creators
Every aspect of our business is designed to improve another, and everything is aligned so that the number of blockbusters continues to rise.
If you're interested in building at the intersection of ai, tech and entertainment, DM me.
There we go.
Never be overly doomeristic
This post looks like the start of a VERY sophisticated and well-funded PR operation to get support for Democrats to regulate AI into oblivion. Let me show you how it works:
1.) This guy, with minimal followers and no previous account activity, goes to the Wall Street Journal which publishes an exclusive with quotes from him on his resignation 18 minutes BEFORE this post goes up. Planning was clearly done in advance.
2.) Within hours, it has tens of thousands of reposts and the account has 100k+ followers. The post is punchy, quotable, it almost seems professionally written. The first three accounts to quote tweet it all do so within 15 minutes of the initial posting. Remember, this account had basically zero engagement beforehand, so an organic reach explanation seems unlikely.
According to Grok those accounts are @_NathanCalvin (General Counsel at Encode AI), @peterwildeford (Head of Policy at the AI Policy Network), and @DKokotajlo (Head of the AI Futures Project), all of which are up-and-coming AI-Doomer policy advocacy nonprofits.
The AI Futures Project website says it is funded “primarily” by the Survival and Flourishing Fund, which says on its own website that it has advised Jaan Tallinn, Skype creator and one of the leading investors in Anthropic, to grant over $2.5 million to the AI Futures Project since 2024.
Encode AI says on its website that it is ALSO funded by the Survival and Flourishing Fund, which in turn says that it told Anthropic investor Jaan Tallinn to grant $516,000 to Encode AI in 2025.
And wouldn’t you know it, the Survival and Flourishing Fund ALSO says it told Jaan Tallinn to grant $2 million to the AI Policy Institute, the 501(c)(3) affiliate of the AI Policy Network, as well.
What are the odds that the first three quote tweets of Coxon’s post would all be major AI-restriction policy advocates funded generously by the same donor, who also happens to be one of the leading investors in, and a board member of, Anthropic, the company Coxon was resigning from? And all within 15 minutes of posting (two within ten)?
3.) Jacob Coxon doesn’t have much of a resume, but we do know that, in 2022, he got a $20,159 scholarship for the “long term future scholarship program” from the Good Ventures Foundation, one of the philanthropic vehicles of Dustin Moskovitz, a notorious AI-doomer who has spent tens if not hundreds of millions on policy advocacy to strictly regulate AI, while also being an Anthropic Investor himself.
It also just so happens that the 14th person to quote Coxon’s post was @MaxNadeau_ (27 minutes after posting) who is the program officer for the Technical AI Safety team at Coefficient Giving, another of Moskovitz’s philanthropic spending vehicles. Max is not a frequent poster, his last posts before quoting Coxon were before Labor Day, but he was remarkably quick off the mark for this one.
4.) Basically every major Democrat politician and candidate has suddenly glommed on to this post, and conveniently, as the people cry out foe answers, Bernie Sanders already has a bill written to “ban super intelligence” and regulate AI into oblivion, and will be releasing later this week. The bill, among many other things, will create “a new cabinet-level federal agency to safeguard the public from the dangers of artificial intelligence” that will be “advised by an Artificial Intelligence Advisory Board comprised of experts on artificial intelligence.” Do you think, perhaps, Anthropic and its many investors who fund AI policy advocacy might have interest in getting to place a pet “expert” on the board of an entity that dictates what AI is and isn’t allowed to do? And isn’t it fortuitous that this whistleblower came forward with his oh-so scary stories so close in proximity to the release of the most radical piece of AI legislation ever introduced?
Light retweeted
Today I had a call with Meta's global creative strategy lead
Was actually very useful I have to say
anyway, here's something she gave me that I thought I may as well share
Pulled from the top spending global ad accounts:
Summary of the hooks from the top performing ads on the platform
Light retweeted
i used to think consumer apps were dead.
why would i need an app to book a flight if i could just text an AI to search and book on my behalf?
it seemed so obvious that everything i use my computer for could be done with text or voice. this was a hill i was willing to die on.
i've completely changed my opinion.
the problem is that a conversation is a terrible interface for everything.
chat apps are a UI paradigm shift that genuinely feels like a miracle. you can give an app a messy, multimodal input: say something, type something, take a photo.
and the app turns that into intent. it understands you.
but the output is more language for you to consume, correct, and steer.
it turns intent into more work. every text response creates another job and a decision.
the best consumer apps have spent decades learning how to ask the user for less.
but a chat asks for more.
imagine using uber in a chat:
"where would you like to go?"
"i found several possible destinations."
"would you prefer the cheaper ride or the faster one?"
"your driver is seven minutes away. would you like me to proceed?"
uber is clearly a better product because it compresses all of these decisions into a single question:
"where are you going?"
we've confused intent with a conversation when the output should instead be, like uber, the desired outcome.
a calorie tracker shouldn't talk to you about your lunch. it should log it.
a shopping app shouldn't give you a paragraph about what to buy. it should build the cart.
a travel app shouldn't interview you. it should give you the itinerary.
before ai, software had to anticipate how users would express intent: buttons, forms, menus, filters, queries.
now the input can be literally anything.
the next wave of consumer apps is what happens after your intent is understood.
It’s incredible how far you can push the UI with Fable 5.1.
This is an interactive children's book that reads the story aloud, uses layered 3D illustrations, and animates every word.
Fable created the concept and story from scratch based on a prompt. You can reuse the result as a template for more children's books, then have Fable voice, animate, and illustrate each one.
Light retweeted
Our new image model generates images 2x faster than GPT-Image-2, currently the best model in the world.
It's also 72% more efficient in GPU usage, so we can provide it at an incredible price.
This gives it the best price-performance score in the world.
Unbelievable work from the team. So much more to come! Try MAI-Image-2.6-Flash out now!
Light retweeted
TikTok all public endpoints reversed engineered. Scraped 5.94 billion TikTok videos and 3.23 billion profiles in 3 weeks. tiktok-api.seeksocial.io
Muse beats sol and opus??
Things are about to get exciting
Light retweeted
if your "marketing engineer" doesn't know what to build first
please for the love of god fire them because they're fleecing you
or send them this so I don't have an aneurysm
everything below is just agents, agents are just code with a thinking loop, connected to a live data stream they are making decisions on
to do everything below you need to give you coding agent access to the following:
- Data pipeline
- Data warehouse
- Cloud Server
- Media Storage (Images, Videos)
- Databases for agents (Postgres databases)
- Recurring tasks (Cron jobs)
- Application Authentication
- Sharable Links
- Git origin for multiplayer access
- API gateway (Nano banana, Apollo IO, Seed dance, Apify etc)
paid ads
google ads
find bottom of funnel keywords related to the product using data for SEO API and a coding agent then build a campaign that's optimizing for a conversion action as deep as you can go within the funnel structure your campaign with as many ad sets as you can ad sets are keyword families the keywords the ads and landing pages all align negative match any keywords that you show up for that don't produce results or don't have a search intent that's related to the product use an LLM to judge search intent
facebook ads
Figure out the desired outcomes of your target customer by scraping social media Then build ads around these desired outcomes present your products as the ads Make statics and UGC, they both work And then structure the campaign like this one Advantage+ campaign, broad, CBO, optimize for the deepest event that gets ~50/week, and spend all your time making more creative
linkedin ads
scrape linkedIn creators content within your category and remix their content into your own then publish that content and if the content already went viral, your remix version will also go viral then use that post to do a thought leadership ad have a call to action within the thought leadership ad to go and sign up for the thing and anybody who engages with the thought leadership ads, we're going to save those for cold outbound below and optimize for conversion action as deep within the funnel as you can go
SEO
traditional SEO
find bottom of funnel keywords that are related to the brand X vs Y, X alternative, X review, how to X that you integrate with then research what is ranking on page one for those keywords extract that right in article that's based on this put your own spin on it by providing a transcript of your unique point of view publish this article, refresh this article monthly
AI search
I have something to tell you AI search is just SEO, but the best content for this is best X for Y AI search sources, the results from page 1 through 3, you need to show up there how do you show up there by doing traditional SEO
Link building
traditional SEO needs link and if AI searches is just traditional SEO it needs links How do you build links? Find a piece of content that a lot of people are linking to. Then write your better version of this, then cold outbound to people that link to that old content that is stale. And say, "Hey, I wrote a better piece of content than the one that you're already linking to on this page here." Can I get this link pasted because it's a better piece of content? And a certain percentage of them will update
Cold Outbound
cold email
For cold email you need to extract engagers from social posts on LinkedIn people who have engaged with content that would want to buy your product then do a waterfall enrichment through multiple data providers typically three to five you're gonna find emails and you can do all this with Apify from a provider called App Maestro and the waterfall enrichment through something like Git leads to find email to Apollo validate the emails with million verifier add them to instantly for the inboxes use something like hyper tide I owe you can send 10,000 cold emails for about a hundred dollars a month
Light retweeted
This is crazy. This Minimax H3 finetune can generate videos in realtime (15s in 13s), up to 14x speedup
Fully open source with training code coming soon
huggingface.co/FastVideo/Fas…
