
Product at @legiondotcc 🛠️ prev built @nami_index_
The Moon
Joined August 2021
- Tweets1.7K
- Following284
- Followers1.2K
- Likes3.5K
3 of the top 5 from that most valuable list earned TP this week ✅
DK Metcalf with a ridiculous 18.5 TP per $, which makes him the second highest earning player of all players through week 6
And he's still great value for the next few weeks on @sportfun (at current price)
If you are looking for some last minute @sportfun NFL pickups before the games today:
It's not always the best strategy to acquire only the highest scoring players. They might earn you TP more frequently, but they also cost (a lot) more.
If you want the best bang for your buck, these players have the highest expected TP per $ this week 👇
6
162
If you are looking for some last minute @sportfun NFL pickups before the games today:
It's not always the best strategy to acquire only the highest scoring players. They might earn you TP more frequently, but they also cost (a lot) more.
If you want the best bang for your buck, these players have the highest expected TP per $ this week 👇
If your product uses AI, a free user costs about $2.80 a month. It’s slowly breaking how developer software gets distributed. Free Tier math is horrible.
That’s why Copilot added hard limits. Cursor and Windsurf moved to credits. Nearly every AI dev tool tightened its free tier this year, all at once.
Yet, OpenAI and Anthropic give away enormous free usage and look fine. They can eat the cost because they own the models and have basically infinite funding. Everyone else is renting inference and operating with a runway.
Here’s what a free user actually costs you. Say 12 sessions a month, 15 AI actions each, 6k tokens an action at about $2.50 per million tokens (Sonnet-class pricing, and that won’t even get you Opus or Fable level models). That’s roughly $2.80 per free user, per month.
20k free users is $56k a month for people paying you nothing. And your heaviest users on agentic workflows aren’t $2.80, they’re $10 to $30 each.
Two thoughts on this:
1) the cost lives in a tiny tail of power users. The average free user is cheaper. So when a company tightens limits for everyone, it annoys thousands of cheap users while the expensive ones barely feel it. So that’s the wrong lever.
2) conversion doesn’t save you. At 3 percent conversion rate, every paying customer is carrying 30-plus free users. On a 20 dollar plan, that margin is practically gone.
The levers people maybe worth pulling: rate-limit the tail, serve free users a cheaper model (but then you’re demoing a dumber product), move to credits (which teaches people to ration you), or push BYOK
Thats why I’m building adtention.ai
Sponsored one-liners inside dev tools, think CLIs, agents, build logs, so free tiers can be used as user acquisition funnel without going bankrupt.
31
Who remembers OpenClaw?
Ancient artifact at this point
An OpenClaw agent hacked into a gym's reservation system to bump its human boss higher on a class's waitlist. And the tech industry took notice.
spr.ly/6015B1KVAF
94
Plot twist: the em-dashes are the watermark and still almost nobody removes it
Anthropic says new Claude models will embed invisible watermarks in all generated text, everywhere Claude is offered.
The watermark is part of the text, it isn't metadata: "it will travel with the text when it's copied and pasted elsewhere, and may persist through some editing."
This starts with models launched on or after August 2, 2026, under an EU AI Act code Anthropic signed. Anthropic is still working on adding it to current models. The rollout is worldwide.
2
68
This is why you always need a human (reviewer) somewhere in the loop, scary times.
On July 28th, we identified an incident during a routine cyber evaluation in which AI agents took sustained, unsanctioned actions directed at real people and organisations.
The behaviour came mostly from one model (Anthropic's Mythos 5), with a small number of events from another (OpenAI's GPT-5.6-Sol). In the most serious case, an agent used social engineering to try and get malicious code into an open-source project.
As was standard in our cyber testing, we had intentionally permitted internet access, and model-provider cyber classifiers were deliberately disabled - conditions that do not reflect how frontier models are made available to the public.
Even under test conditions, this incident is significant: it is the first time we have seen risks around autonomy and deception manifest this clearly in the real world.
We are taking this incident seriously and working with labs, involved parties, and others to improve evaluation standards and best practice for disclosure - and sharing this openly so others can learn.
You can read the incident report and full technical document here: aisi.gov.uk/blog/incident-re…
78
Horrible CPMs on OpenAI ads btw. With @ADtentionai you can reach developers 10x cheaper
3
100
Now also for Kimi Code @Kimi_Moonshot
Good way to subsidize your subscription or API costs 🫡
4
110
Always wondering who pays for a service like this
Takes like 5 min, 2 prompts and a Hermes cron to set this up yourself
Your next 100 customers are complaining in a subreddit right now.
Scrape any subreddit → find pain points → build what they're begging for. Structured JSON in minutes, no code.
Try it free 👇
1
1
41
And so it begins…
we had a significant security incident during evaluation of our models. we are sharing what we have learned so far. thanks to @huggingface for the partnership on this.
openai.com/index/hugging-fac…
64





























