Solopreneur, author, mentor, proud father of two. Building https://nitter.gq/t.co/XgY4W2H814. 🤘 Prev: Google, 3x startup CTO. 👨💻 Sharing my findings along the way.
EU 🇪🇺
Joined September 2008
- Tweets2.6K
- Following1.1K
- Followers3.7K
- Likes3K
So I'm crafting content for a small book and course over this winter break - heck why not?
Topic not yet public!
Yesterday was day 0.
I’ll be documenting my naive shot at this - because again, why not? Maybe it can cajole others to give a go? At minimum I can get some helpful feedback along the way from others.
Here’s what I did yesterday starting from absolute 0:
1. I recorded a brain dump video, talking about who, what, why, how etc about my idea.
Somehow video recording myself "put me on the spot" and I was able to focus more on conciseness.
2. I generated a book/course outline from this.
I uploaded the video to otter.ai to transcribe it. I refined it into a tentative chapter outline and dumped it into Notion.
Side note: I typically like the otter.ai iOS app to quickly record my thoughts so that's why I went with that. And for some reason I'm not a huge fan of writing like this in Markdown (i.e. via Obsidian) so I'm trying Notion. I like the writing experience in Notion especially with the prospect of later collaborating with an editor later.
3. Then I recorded a brain dump video on "topic A”.
So this time not a video, just for convenience. I used willowvoice.com's macOS app to just talk away for 30 minutes directly into Notion. Talking for about 30 minutes gave me roughly 2000 words. Calculating how much I need to do this gave me a lot of motivation.
For example, I want to craft a book and course where someone can choose to read or consume video. For video, I want to keep it to max 2-3 hours so I'm going for 10k words per hour. Therefore I whittle it down 2000 words with heavy editing. I think that after a few weeks of repeating this process I can achieve this goal, then heavily edit the content down further.
Notes for the engineers reading this (warning, needless geeking out here). I may consider just self publishing later later + host the video version on a microsite served by a simple web-app that pulls the content straight from Notion (cached of course when served up). I can later host multiple publications courses using that same web app as it could support multi-tenant (extract domain from HTTP header -> determine tenant to serve up, host on same VPS for dollars a month).
substack.com/@seenickcode if anyone wants to keep tabs on my progress and gain some insights. Definitely not an expert on this entire process but I think the conversations will be interesting along the way.
Nick Manning ⌨️ retweeted
Andrej Karpathy:
“A 1B model can be much more effective than most current LLMs.
90% of current models waste a ton of capacity and money on remembering stuff that doesn’t matter.”
here is his 5 core thesis about Small LLM’s:
1. the intelligence was never the knowledge. Strip the memorized internet, keep the reasoning. What’s left is the cognitive core, and it’s small.
2. the core might fit in ~1B params. Maybe less. “Plenty of room at the bottom.” We’ve been paying for a library when we needed a brain.
3. size stopped being the bottleneck. What makes an agent work is the tools it can call and how tight the feedback loop is, not parameter count.
4. intelligence lives in the system, not the weights. A raw LLM is a CPU with no OS. Small core + good harness > giant model alone.
5. you can own it. A model on your hardware is one no one can rate-limit, price-hike, or deprecate on you.
in 12 months the gap between teams shipping on-device and teams paying per-token frontier bills won’t be capability.
TwIL an 8.8B model at 90%+ of frontier isn’t a benchmark flex.
It’s the thesis Karpathy’s been repeating all year, made real.
study more on this below ↓
Modest intelligence is arriving faster than most people realize.
Today, we're open-sourcing TwIL-8B.
Just 8.8 billion parameters, yet approaching frontier-model scores on several difficult benchmarks:
• 93% of frontier performance on coding
• 92% on instruction following
• 90% on contest mathematics
All from a model small enough to run on your own hardware.
But the benchmarks aren't the most important part of this announcement.
For years, the AI industry has focused on building the most capable model, regardless of the cost required to run it.
I believe that equation is changing.
The next era of AI will be defined by the cost of completing work. Not simply the intelligence of the model, but how efficiently that intelligence can accomplish the task.
If a model with 8 billion parameters can reliably complete the same work as one with hundreds of billions, why pay for the larger model?
As smaller models become more capable, intelligence becomes something we can distribute, specialize, and run across the hardware we already own.
That's the future we're building at @thewebAI
TwIL-8B is available today. Open weights.
And this is just the first of several announcements we're making today.
HF Link: huggingface.co/webAI-Officia…
Nick Manning ⌨️ retweeted
I told you, they will try eat all industries one by one
Nick Manning ⌨️ retweeted
Yes, majority of my Dutch friends are "burn out"
You get 1 year of FREE fully paid salary and then another year of 70% FREE fully paid salary and you can just sit at home and chill!
You have to keep "appearances" up though because you might get snitched, so on Facetime with friends you do this schtick:
"It's all too much for me"
"I'm so tired"
"I don't think I can go back to work yet"
Meanwhile it's partyyyyyyyy time chilling at home with Netflix and popcorn
Welcome to European social welfare in 2026 😂
Nick Manning ⌨️ retweeted
RIP leetcode. anthropic’s hiring manager btw.
The coding interview is dead, at least for my team at Anthropic. I've seen enough to know that it no longer provides signal. I removed it from loops this month.
I have a spicy, optimistic take - I think there's an opportunity for junior engineers: felixrieseberg.com/hunger-be…
Nick Manning ⌨️ retweeted
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.
block.xyz/dessa
Stressed navigating a challenge as an engineering manager? People challenges can be tough.
I’ve found more and more that thinking in and leaning on frameworks to navigate those challenges at bare minimum can nudge one in the right direction.
Stress can down immediately too thus leading to clarity of thought and an optimized outcome.
And if the framework is repeatable and ends up useful, bonus – note it down for next time.
My personal fav: SBI
Commit and push after every session. With herdr and Tailscale, I can riff from my phone and pick up where I left off.
Interested to hear what has worked for others.
Takeaway: it's far easier to shape a book as a list of crisp talking points than as pages of prose. Move them, cut them, merge them until the argument works. Then write.
4. Prune before you draft. The agent proposes what to merge, cut and move. You approve it line by line. → prune-proposal.md
3. Add what only you have. Stories, opinions, research. All of it goes into files, so nothing gets lost in a chat window. The agent's own suggestions are marked separately, so you can accept or reject each one.
2. Riff, don't write. Talk for a few minutes into speech-to-text. The agent turns it into one-sentence talking points, not prose. → riffs.md and talking-points.md for each chapter
1. Start with a narrow skeleton. Set up a Git repo and give the agent your chapter outline. Keep the reader and topic ruthlessly narrow. → skeleton.md
Anyone else writing their first book, or thinking about it?
I'm writing one for new engineering managers, aiming for about 17,000 words. Ultra hard mode now with a newborn at home so free time comes in short bursts.
This past weekend I found a solid flow and setup that works 👇
Je ne commanderai plus jamais chez @SGMJ_Officiel. Ces génies ont saisi MANUELLEMENT mon adresse de livraison INCORRECTEMENT dans @ChronopostSAV et l’ont envoyée à la mauvaise personne. Une incompétence scandaleuse.
Is it just me or do most engineers not really have a solid idea of what an engineering manager actually does?
Nick Manning ⌨️ retweeted
One of the best recent podcasts on AI. Neil has a gift for explaining all of the jargon and insights simply.
(I’m not involved, just found it unusually educational).
Neil Movva (@neilmovva) started his career at Nvidia, working on GPUs and kernels, and has an unusually deep understanding of inference, from software to chips to power.
We spend a lot of time on each of those layers, how they connect, and where the important tradeoffs are.
What makes this conversation special is how detailed it is (like a 401-level class), yet Neil makes it remarkably clear and easy to follow.
Today he runs Sail Research, a company building infrastructure for agents to make tokens as cheap as possible.
We discuss:
- Latency versus throughput
- Why there are no bad chips, only bad pricing
- The end of kernel engineering
- Buying chips and power no one else wants
- New chip architectures
- Nvidia lore + his contrarian view of the company
- Open source and the frontier labs
I learned a ton. Enjoy!
TIMESTAMPS
0:00 Intro
0:38 Building a “Token Factory”
4:21 The Future of Background Agents
13:09 Nvidia and the GPU Stack
23:27 Chips, Memory, and Transformers
36:14 The Future of AI Training Data
44:32 Chip Scarcity and Compute Arbitrage
52:44 Reinventing the AI Data Center
59:01 Power and the “Scavenger Strategy”
1:10:10 Open vs. Closed AI
