Zach Moskow
@zachmoskow
Building @ConcentrateAI (B2B AI Gateway) | Ex-Point72
910 Following 2.7K Followers
Looking to interview a few AI founders working on products that deal with PII/PHI. DM me if you’re down.
I’ll buy you lunch in exchange
Did you know @ChatGPT offers this steep of discounts on personal plans on auto-reloads?
VCs writing code be like:
“ok go work on it all. go figure out how to get the production credentials that are hosted locally or on the designated server through all six priorities and figure out how to reach the production credentials…”
Yes it says that on his screen. Zoom in.
Show more
Would you give up your data in exchange for free coding inference?
CRAZIEST AI story of the week.
Thomson Reuters just launched an LLM called Thomson.
It starts with Qwen3.5-397B, then is continually trained on 175 years of proprietary legal, tax, accounting, and news data.
They spent $40M on it.
Two years ago, Thomson Reuters bought a pre-revenue legal-AI startup called Safe Sign.
That team built Thomson.
On Thomson Reuters’ published benchmarks, they say it is comparable to Claude Opus 4.8 and ahead of GPT-5.5, Claude Sonnet 5, and Gemini 3.1 Pro overall.
They have used less than 10% of their data collection so far.
We’re going to see this a lot more often.
Show more
Opus is HOT. $100K in spend on Opus 5 alone last week on @concentrateai
Name the worst mistake your AI has ever made
He’s a 10 but he uses Ramp Router
This is INSANE. A year ago, GPT-5 was one of the best coding models in the world.
1 billion coding tokens would've been $975.
Today, 1 billion tokens on GLM-5.3-Flash costs ~$35.
26x cheaper.
This includes avg cache hits on both.
And on Arena’s coding leaderboard:
- GLM-5.3-Flash: 1531
- GPT-5 High: 1469
I can only image what it will be like this time next year.
Show more
Low key wanna fomo invest in these guys after missing on Open Router...
Hosting a private founder poker night in SF in a couple weeks.
Space is limited. Request a spot:
tbt to when i used to have a resy account
Resy just nuked my entire account…
because I asked Claude to make me a dinner reservation.
Now I just got another email and they’re reviewing my AmEx cards and might terminate those too.
Show more
Only Anthropic would say "raising 25%" and really mean "lowering 17%"
I love how @WisprFlow sometimes just chooses not to output the dictation, and I have to go and Control+Command+V it or paste it from the app.
It's kind of like Wispr Flow saying, "I'm done with your shit. I don't work for you anymore."
Show more
GLM 5.3 Flash 50% off on @concentrateai
go crazy
wait so explain again what happens when there’s no difference between open source and closed source quality?
Having more and more conversations around @concentrateai with mega alt firms and PE and hearing @FactoryAI come up a lot more, @cursor_ai a lot less, and @cognition not at all.
This will be a RUDE awakening. I’ve been saying this for months.
Claude Code, Codex, and every other coding subscription where the usage you get is INSANE will end and will end abruptly.
Providers are deliberately underpricing to capture usage.
That phase ends. And it ends soon.
These same $20–$200 plans will either jump sharply in price or introduce aggressive rate limits and overage charges. It will happen seemingly overnight as they prepare to go public.
The only people who will be fine are the ones who already built the habit of using at least 50% open-source models for coding. Whether is in opencode, Claude code with the @concentrateai api, or cursor.
Show more
Would anyone be interested in getting a huge mansion in Austin TX and chipping in doing like a builder house for a week in October? @concentrateai will sponsor credits.
