GTM | AI Enablement | Everything is under control
London, UK
Joined April 2010
- Tweets5.1K
- Following603
- Followers1.1K
- Likes161K
Dylan Lamb retweeted
What separates the winners from the losers is how a person reacts to each new twist of fate.
Dylan Lamb retweeted
Just find "the Button"
Every market's prospects have one.
You just gotta find it.
And you gotta make sure it's the MAIN button
The one the majority of that market has.
it comes in 2 parts:
1. {common enemy} is purposely keeping you {broke|fat|lazy|sick}
^ this validates their failure trying to solve it themselves already
2. There is a simple trick they don't want you to know that fixes it
^ this gives them hope
The key is getting them to believe part 2
They WANT to believe part 1. It's their scapegoat. We all know how well "it's not your fault" works.
But make them believe part 2, and you can press the button over and over and over and over and over again and literally print money.
It's all about the positioning of the message.
The product team at Gemini should focus on building around education and sciences. That's where Argon shines the most from my experience with it.
their blood must be boiling
JUST IN: French blood bank hit by cyber attack with blood samples being heated to 110 degrees Celsius from their maintained -30, per Le Canard Enchaîné.
Readers added context they thought people might want to know
The 110 °C figure was the value hacked equipment was reprogrammed to reach, which does not mean the cold rooms or plasma bags actually reached 110 °C. France's health ministry said the anomalies were spotted in time, with no operational consequences. cyberattaque.org/cyberattaque-c…
Dylan Lamb retweeted
ChatGPT and Codex are helping engineers build for the physical world.
We’re sharing our progress across mechanical design, electrical engineering, and simulation—from helping our hardware teams iterate faster to testing new workflows with industry collaborators.
Dylan Lamb retweeted
maybe a hot take but i now consider adversarial reviews mostly a waste of tokens. far better to use those tokens on empirical combined with formal verification. i have updated pstack to use the interrogate skill less, and eventually i think i may remove it altogether
i have found very little value with code review in the abstract, agents tend to just get lost in the nitpicks and yagnis
Dylan Lamb retweeted
this is basically the bitter lesson at work
the smartest models will need fewer and fewer tokens, they won't need to verify their work or check it for bugs, they will just be right
maybe a hot take but i now consider adversarial reviews mostly a waste of tokens. far better to use those tokens on empirical combined with formal verification. i have updated pstack to use the interrogate skill less, and eventually i think i may remove it altogether
i have found very little value with code review in the abstract, agents tend to just get lost in the nitpicks and yagnis
the speed at which Varick is growing is pretty insane. i still remember when vas was doing vsl's summer last year
Hiring a Head of Marketing for Varick.
In the last year we've done over 300M impressions on socials, without even trying. It was me posting between meetings or while walking to work.
I want you doing this full time so we hit 3B in the next year.
Must be chronically addicted to AI Twitter, and must have S-Tier video/image editing and design experience.
Up to $200K base + equity. In person at our SF HQ. DMs open.
Dylan Lamb retweeted
For anyone questioning whether this post is overhyped, I think they are actually *underselling* the approach.
We did similar experiments and found that Claude models can write entire, functional neural net weights from scratch! csinva.io/weight-space-progr…
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
Dylan Lamb retweeted
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
Generational company and that's without even addressing their line about RSP
The most important takeaway from
the 3Q numbers is that Anthropic’s revenue per $1 of compute declined from 2.08x last quarter to 1.72x this quarter
As reminder, the second quarter of 2026 was the peak moment for token maxing, and also when every company began to onboard Claude Enterprise
Believe the compression in revenue per token is coming from a combination of
> Enterprises pushing spending limits on employees, resulting in employees using cheaper models to do basic tasks
> Optimization tools such as OpenRouter pushing spend towards the most efficient models for specific task
> Prior quarter likely benefitted from people experimenting with Claude due to the halo effect of trying a new technology
Either way, overall numbers remain solid and growth is still showing on top line. Cash balance remained relatively steady, which I thought was impressive
This is clearly a “once in a generation” type of business, as you can already tell from the valuation it is going public at
Dylan Lamb retweeted
The most important takeaway from
the 3Q numbers is that Anthropic’s revenue per $1 of compute declined from 2.08x last quarter to 1.72x this quarter
As reminder, the second quarter of 2026 was the peak moment for token maxing, and also when every company began to onboard Claude Enterprise
Believe the compression in revenue per token is coming from a combination of
> Enterprises pushing spending limits on employees, resulting in employees using cheaper models to do basic tasks
> Optimization tools such as OpenRouter pushing spend towards the most efficient models for specific task
> Prior quarter likely benefitted from people experimenting with Claude due to the halo effect of trying a new technology
Either way, overall numbers remain solid and growth is still showing on top line. Cash balance remained relatively steady, which I thought was impressive
This is clearly a “once in a generation” type of business, as you can already tell from the valuation it is going public at
unreal
Anthropic says in its amended S-1 that its internal estimates and expectations about FUTURE models indicate that it may "imminently" meet the definition if "dramatically accelerated progress attributable to AI".
Under the RSP, this means that Anthropic observes or expects 2x the rate of progress in AI capabilities vs. the rate Anthropic would expect and the fastest rate of extended progress observed over at least 3 model generations without significant AI contributions to AI R&D, plus it's plausible that this doubling is due to automation of research and/or engineering such that it's likely that there will be even greater acceleration.
Note that "2x the rate of progress" means that 2 years of progress would be compressed into 1 year. So, if Anthropic would have improved its models 9x in one year without AI-enabled acceleration, this threshold would be reached when Anthropic expects to be able to improve its models by 81x in one year (9^2).
Dylan Lamb retweeted
Data is becoming a commodity. It's a race to the bottom price-wise nowadays.
For example, we just came up with a solution where we can have THE FRESHEST DATA ON THE MARKET, all the time, for $70-80K/mo.
If you are a data company that's betting only on the data that anyone has, you don't have a MOAT on the long run.
This doesn't mean that you can't sell to the 98% of the market that's still not aware of this, but I'm seeing more and more companies solving it super easily and super fast (eg: @blitzapi, LeadOcean, Moltsets, etc.).
The real MOAT is the distribution, for now. If you are a data vendor that has nailed the distribution and you are present EVERYWHERE, you'll probably win for a few more years. Because the market adopts slowly.
We live in the bubble, and you see the evangelists like Growth Engine X, Taylor, Balsa, Understory adopting these things immediately, but for the rest of the market (which is HUGE, and probably 98%), it's still "let me buy ZoomInfo/Apollo/Lusha"
The next data moat is data generated from leads' behavior.
- Is this lead more likely to reply to LinkedIn DM or an email?
- What's the phone pickup rate of this lead?
- How is this lead replying to this kind of offers?
- Which job title or industry is this lead most likely to reply to?
This change will come slowly, but it's coming for sure.
If you plan to be relevant in 3-4 years from now, you have to adapt.
Dylan Lamb retweeted
Too many people make the mistake of conflating their occupation with their purpose. Your ‘purpose’ is to grow, develop, and evolve, as a soul through incarnational experiences to develop wisdom through navigating life and mastering energy.
Do not confuse THAT, with your JOB.
Lmao no way
I recreated subvocal in 9 minutes with one prompt for AirPods Pro 3.
Source: github.com/joeblau/subvox
