Been working as a dev for 5 years | Posting about tech and AI every day (at least i try to lmao)
Boston, MA
Joined March 2015
- Tweets370
- Following123
- Followers69
- Likes690
Live web search with actual citations is now available for US government workflows thanks to Perplexity’s new FedRAMP compliance
2
𝚁𝚘𝚗𝚊𝚗 retweeted
anthropic committing 5% of its shares to a new public benefit fund is a cap table decision before it's a pr one.
that's a large slice to carve out, and founders should notice the order: governance structure first, narrative second.
my take: if your startup makes a mission claim, write it into the equity structure early. investors are going to price it either way, so you might as well be the one who set the terms.
banks are experimenting with AI financial assistants and i just know mine would spend 10 minutes analyzing my transactions before asking if everything is okay at home.
3
Trying to resolve a real issue with an automated customer support chatbot is peak modern frustration. The bot asks five useless questions just to link you to a broken FAQ page.
5
the next billion-dollar startups probably won't look like another productivity app
they'll own boring industries
logistics
manufacturing
insurance
energy
the less glamorous the workflow, the bigger the opportunity might be
9
Spending $1500 on a laptop with an integrated "AI NPU" just so Windows can index your screenshots in real-time is peak tech consumerism.
9
𝚁𝚘𝚗𝚊𝚗 retweeted
OpenAI citing past contract violations by Musk's other companies as the reason for ending the Cursor partnership is a much more pointed explanation than a typical business disagreement
it also suggests that this decision was less about Cursor and more about OpenAI drawing a line against doing any business with SpaceX affiliated entities
Musk calling Altman untrustworthy in response confirms this is personal as much as it's commercial, adding another layer to their beef 😂
moonshot's tech blog explicitly says K3 "still trails" fable 5 and gpt-5.6 sol overall, while topping every other tested model. "best open-weight model" is accurate; "big AI is scared" oversells where it actually sits relative to the two closed frontier leaders.
I've been testing Kimi K3, and holy smokes, this is the best open-weight model the world has seen.
It's a 2.8T-parameter vision model that supports tool calling and reasoning, with a 1M context window.
And it's really good.
I now understand why Big AI is scared.
26
the biggest risk here is discovering the emergency buffer isn't nearly as usable as the headline number suggests when you actually need it.
america's oil buffer isn't just shrinking, the mechanism behind it matters more than the headline number.
worth adding - the Cushing, Oklahoma delivery hub, the actual settlement point for NYMEX crude contracts, has already dropped below its ~20 million barrel operational minimum. that's a physical logistics constraint, not just an inventory statistic - it affects how crude actually gets delivered and priced, separate from the broader reserve story.
context that matters: the SPR drawdown is structured as an exchange (companies borrow crude, return more later with an 18-28% premium), not a straight depletion. energy secretary wright has framed this as "for every barrel released, we get back more than 1.2 barrels next year." the real risk flagged by analysts isn't the mechanism, it's timing - "how the oil market balances if the SPR runs out before peace and stability arrive."
worth the caveat too - GAO already found the SPR can only draw down at about 61% of its designed rate due to aging infrastructure, meaning the reserve's actual emergency-response capacity is more strained than the barrel count alone suggests
1
8
the "own the app layer, give away the model" logic tracks with the cursor acquisition and grok's existing infrastructure buildout. the space data center part is a much bigger, more speculative leap though.
Elon’s original vision of an open, American AI was right.
He could still flip Grok to open source which could be checkmate.
Why?
If you flip Grok open source, you necessarily move all margins out of the model layer and down into the infrastructure (chips + CSPs) and above to apps.
He owns a massive data center build, owns next gen inference AI and owns the most used app (Cursor). So arguably he would make *more* margin by doing this. Then moving the DCs to space makes even more sense.
11
same-day availability for gpt-5.6 terra and sol is a strong signal of how fast perplexity ships new frontier models to users. no waiting around here.
20
the new investor relations playbook writes itself.
forget the earnings call.
just announce a $250M contribution and let the President do your PR.
BREAKING: President Trump says Micron stock, $MU, went up “9 points” after the company announced a $250 million contribution to Trump Accounts.
17
half of tech twitter's problems would disappear if people spent as much time shipping as they do posting.
3
if a model behaves differently based on who it's helping, transparency isn't optional. users need to know where the guardrails are.
That was quick: Anthropic reversed a controversial policy that would have secretly degraded Claude Fable 5 for users doing frontier AI research after backlash from researchers who saw it as covert sabotage of competing AI development.
1
400bps lag. €10B deployed.
same robots. different timezone.
$AMZN just committed €10B+ to automate European fulfillment. Everyone's talking about the 25,000 jobs. The margin story is what actually moves the stock.
Amazon's North America retail operating margin hit 6.9% last quarter. Two years ago it was 3.5%. Robotics is the primary driver - Proteus, STARK, Deepfleet AI coordinating over 1 million robots across fulfillment centers.
European retail margins still lag North America by ~400bps. This investment is the catch-up play. Same playbook, different geography.
The robots deploying across European sites:
• Proteus - free navigation alongside humans, no safety cages
• Vulcan - touch sensitivity for handling fragile items, coming 2027
• STARK - tote and trailer unloading, historically the most labor-intensive step
• 50,000 electric vans globally on top of all of this
The $1B worker retraining commitment by 2030 is the political cover that comes with deploying robots at scale across five countries. Worth noting - not worth overstating.
Look out for the European segment operating margin in Q3 and Q4. That's where this investment starts showing up in the numbers.
4
Russell inclusion can absolutely create short-term reflexivity.
but passive buying pressure is temporary. the sustainability of the move still depends on whether the underlying unit economics are actually improving.
$OPEN Everyone's making the Carvana comparison. Here's whether the numbers actually support it.
Russell 3000 inclusion kicks in June 26. Passive funds are forced buyers between now and then.
The Carvana comp in plain terms: $CVNA was near-bankrupt in 2022, rebuilt on faster inventory turns and AI pricing, then 10x'd in 18 months. Bulls think $OPEN is running the same playbook at $4.75 with a $10-15 target.
The data that supports it:
Gross profit growing every quarter - $57M in Q4 2025 to $72M in Q1 2026
AI pricing engine cutting days-to-offer to under 24 hours
Faster inventory turns mean less capital trapped in homes
The data that complicates it:
Q4 2025 net loss was $1.096B - one quarter
Q1 2026 net loss still $173M
30-year yield at 5.19% - the housing recovery this thesis needs isn't happening at these rates
Index inclusion creates a buying window. It doesn't create a business turnaround.
June 26 forced buying window. FTSE Russell updates May 29, June 5, June 12, June 18. Q2 gross profit direction is the actual signal.
5