Thomas Long (@thomas_long) profile picture

Product @ Navan đŸ‡ș🇾/ đŸ‡«đŸ‡·

Joined June 2009
Photo by Thomas Long (@thomas_long)Photo by Thomas Long (@thomas_long)Photo by Thomas Long (@thomas_long)Photo by Thomas Long (@thomas_long)Photo by Thomas Long (@thomas_long)Photo by Thomas Long (@thomas_long)Photo by Thomas Long (@thomas_long)Photo by Thomas Long (@thomas_long)Photo by Thomas Long (@thomas_long)Photo by Thomas Long (@thomas_long)
Replying to @tanayj
60% at Navan
1
184
Replying to @JaredSleeper
what if you played almost exclusively Madden and Fifa
1
2
346
Thomas Long retweeted
Weekend hack: HackerTube. I was annoyed with YouTube's recommendations so I had Claude filter down 43M comments over 20 years of HN to 6,454 videos that keep being recommended. *Significantly* better viewing experience, ~2 hours with Opus 5.5: hackertube.org/ - Excellent recs without any logins (state is kept in local browser session) - Thumbnail is a representative frame from middle of the video instead of the super duper A/B test hyper-optimized user uploads - Underlying data from the comment analysis as one zip file, point your Claude at it and ask for your own version - HackerTube Longs ✊ A bit more on the process: I use Claude Code (desktop app for Mac), Opus 5.5 on Medium effort as my default. I only talk to Claude and have it do everything, the only "manual work" I helped with in tandem was in the browser (reminiscent of MenuGen): 1) I had to create some kind of API key from Google Cloud UI to query YouTube for video metadata, 2) I booted up a VPS box for serving and 3) I bought the domain name. Once the box was up I gave Claude the IP and told it update, secure, harden, deploy and optimize the code, and it did the whole thing. Many awesome projects that would have normally taken weeks/months even can now be few-shotted for breakfast. What else are we not thinking of.
2,412
4,858
1,825
52,643
5,896,670
Replying to @antoniogm
The best street in SF !!
1
397
Replying to @ShepOgden
đŸ”„đŸ”„
1
119
Replying to @Jack_Raines
bigger than Kraft Heinz !
179
Replying to @kylebrussell
Many Zitron fans expected this S-1 to confirm his POV. Now that it is out, cognitive dissonance will hit, so their first reaction will likely be "there is fraud."
4
17
1,867
Thomas Long retweeted
The most mind blowing stat we've seen so far is this: "As of June 30, 2026, we had raised an aggregate of $141.1 billion of capital and had cash and cash equivalents of $112.4 billion and short-term investments of $19.3 billion." That means to date, they have burned 9.7b, they are growing at a rapid pace, and are profitable on an operating income basis as of the June quarter! This is likely the most efficient scaling of a business in the history of capitalism. xitter.gq/AnthropicAI/status/211

Anthropic has filed an S-1 registration statement with the SEC related to its proposed IPO. Read more: anthropic.com/news/s1-sec Legal disclaimer: A registration statement relating to these securities has been filed with the SEC but has not yet become effective. These securities may not be sold, nor may offers to buy be accepted, prior to the time the registration statement becomes effective. This post does not constitute an offer to sell or the solicitation of an offer to buy these securities, nor shall there be any sale of these securities in any state or jurisdiction in which such offer, solicitation or sale would be unlawful prior to registration or qualification under the securities laws of any such state or jurisdiction.
151
481
123
6,333
866,409
Sales has the GTM engineer. Software has the forward-deployed engineer. Finance is about to get its own. Each of these roles took work that used to need a team and turned it into a system: - GTM engineers automated the outbound, enrichment, and data work that used to take a floor of SDRs - RevOps leaders turned go-to-market into a measured, repeatable process - Forward-deployed engineers embedded with the business and shipped working software without the spec handoff Finance is a natural fit, since the work is high-frequency, rules-heavy, and full of data, but hadn't been leveraged in the same way until recently. Here's what changed: 1. AI agents can handle real finance work, from drafting entries, running variance analysis, and assembling reporting 2. ERP, data warehouse, and bank data is reachable through APIs, so agents can read and act on the numbers directly 3. Finance teams are expected to do more without adding headcount In practice, a finance engineer connects the data, builds and sequences the agents, sets the approval steps, and reviews the output. The finance engineer is the finance function's version of a shift that has already reshaped sales, operations, and software, moving from doing manual work to building and operating the systems that do it. At @Concourse_ai, we build, share, and govern the agents that finance engineers use to run close, FP&A, treasury, and AP/AR. Read more of our thoughts on this shift here: concourse.ai/insights/financ

6
2
5
406
Replying to @nntaleb
il y avait des affiches partout a Paris pour ça il y a quelques semaines !
23
Replying to @PaulSkallas
what do you think of this dream Lindyman? Do you share this dream?
3,050
We own ourselves, therefore we own the effects of our actions. If we own the effects of our actions, then we own that which we produce.
The moral case for capitalism is that your life belongs to you. Your work, your savings and your plans don’t become somebody else’s property because they have a moving speech about how they’d use them.
1
2
406
Thomas Long retweeted
we have millions of dollars to spend on US robotics training data by EOY if you’re a data broker or have relationships with US businesses willing to film their work to train robots, DM me we pay for the data and offer recurring referral revenue, not just a one-time commission huge opportunity here
10
5
32
3,175
Thomas Long retweeted
interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works already-clean.goodfire.com/
890
2,657
1,464
19,996
4,816,620
Thomas Long retweeted
My conversation with @alexandr_wang, chief AI officer of Meta. 0:00 The vision behind Muse 1:32 Discovering the potential of personal AI agents 3:34 Three years of therapy in one AI session 5:46 Building Muse in 7 months 9:16 Why Meta bet on personal AI 16:51 Mark Zuckerberg's obsession with improving products 20:15 How Meta knew Muse was ready to launch 23:32 Why great products need a single point of view 25:56 Making Meta cool again with memes 34:38 Building trust with AI agents 38:43 The future of human ambition 45:22 The call from Mark Zuckerberg that changed everything 52:13 Leaving Scale AI for Meta 55:19 Rebuilding Meta's AI lab from scratch 58:13 From founder to coach 1:05:55 The future billionaires at Alexandr's 19th birthday party 1:08:04 Diamond mining vs. building skyscrapers 1:09:21 Why Alexandr has more than 200 direct reports Includes paid partnerships.
109
254
90
2,414
909,035
Replying to @PaulSkallas
What about warmth and coziness ?
1
12
1,802
I increasingly think the confluence of consultants and AI is “the main show” In 2025 I wrote the “Big Rug” where I argued that AI was going to destroy software and IT services and managed to realize that wasn’t going to happen before the grenade went off The first leg of the re rating was just vibe coding being mostly vibes. The second leg was Snowflake type apps doing incredibly well helping orgs manage complexity. Atlassian was the real face melter - as it showed managers need a harness of harness of sorts bc ai driven orgs are so absurd. The third leg was Claudeforce, and Anthropic hiring an army of FDEs This was a more severe leg - because one of the biggest anti software memes was that every company would customize its own CRM But the fourth leg - is one I find much more interesting. Accenture beat earnings aggressively. Note - this is an org with 800,000 employees. It’s the canonical “AI will replace this” business. Alongside this - you have ppl accusing Cognition of being like Wipro / infosys but meanwhile the money making take is closer to wipro and infosys being closer to cognition You have a slow moving economy filled with boomers who need an ai strategy and you need adults in the room to roll it out or everything breaks in Prod The most recent employment report was telling. Ppl tried to jam the story “AI is halting hiring” but bonds closed red on the day. A single number can’t override the sentiment shift of ppl thinking Accenture would lose half of its employees and then not only that not happening, but hiring expanding there and at salesforce as “enterprise ai” becomes a thing There’s a “Goldilocks ai spend” situation where you simultaneously have everyone spending on inference but also the economy not really decelerating bc people don’t lose their jobs. And any slow down is offset by Trump screaming “accelerate” to make sure his venture investments print. And we are firmly in that - mimetically right now Left curve is Claude force and right curve is the unfalsifiable bull case of Anthropics wet lab or OpenAI solving all math quietly with unreleased models. Mid curve is fading AI or going full Michael Burry wearing his Christian Bale Big Short t shirt like an aging house elf In reality it’s the wrong Christian bale character. It’s more of a Patrick Bateman market where 27 year old sales executives are peptide maxing earning mid 7 figs for selling ai transformations to bloated corporations But behind all of this is the donning realization that Opus 5.5 is a different kind of model. It’s just better. It generates way less code bloat. The Internal CRMs it ships actually work and don’t break in prod. It’s clearly masterful at Tik Tok / marketing. Our feeds went from minimal ai slop to 24/7 just bc it somehow has good taste OpenAI saying next models similar to- don’t bloat as much. In the meanwhile there’s the “cutting my steak with a chainsaw” effect with frontier models. Using fable to make a dashboard is just overkill now with opus 5.5. Glm 5.3 kind of sucks. Kimi is okay. Can’t really use them. But they’re almost there. And are extraordinarily cheap. The 4.1 flash progress from 4 was crazy So despite the market being the most excited it’s ever been about ai Goldilocks going forward we have the most extreme combination of “software being solved” and inference margins compressing just when they need to expand the most (ipo and credit stress) My current mental model is that ai can’t stop bc it’s a national security priority. So Leopold wasn’t wrong - just early. So yes. I am bearish software and consulting here. Even guys like Thiel who spent years saying we were stagnating are now saying we are accelerating. and Labs at some level at $1.5T+ valuations need to do what it takes to win. Which means launching their own software and products w unreleased frontier models. Which the wet labs / math are already foreshadowing $ACN / $IGV lower vs $EWY $NVDA $GOOGL $SFTBY
1
3
62
10,766
Replying to @sriramkri
much worse than Juul
20