Marc Andreessen: Blogposts, Tweetstorms, and Manifestos

David Perell

Abstract

An hour and a quarter on writing as a distribution and thinking practice rather than a craft. The method is fast and unsentimental — write in the voice you talk in, ship before editing stiffens it, and treat the emotional charge as the raw material rather than as something to manage out.

• Input quality caps output quality, which makes the reading and the conversations the actual work.
• The information diet gets barbelled: only the super-current or the timeless, nothing from the middle band.
• A big idea can be a statement of the obvious — the force comes from timing rather than novelty.
• Time is not the binding constraint on writing more; the charge or the idea is.
• A genuinely well-written message gets read, which makes strong writing a standing market inefficiency.
• Write for effect rather than objectivity: feedback that a piece is too polemic can be exactly the point.

Everything below follows the conversation in order, from the emotional charge through drafting and information diet, into group chats, media forms and software as a literary genre, out through style, environment and the big-piece process, and finally into feeds and what a distinctive mind looks like from three sentences. For anyone building a personal brand the operative line is the venue one: publish where you own the audience rather than where you rent it.

Chapter summaries

00:49 - The strongest writing comes out of a genuine emotional charge

The strongest writing comes out of a genuine emotional charge — rage, in one case — with bottled-up energy released onto the page.

"So, so the big one of that it was literally a rage. It was"

— 00:49

Real feeling supplies both the drive to finish and the force a reader feels — one writer described the act as opening a vein. The piece named here was written the night a single absurd news item convinced him something had been lost in the national capacity to build. It is offered as his own pattern, from inside the essay, which he grants is an easier challenge than a book.

03:00 - The voice on the page is the voice in the room, shipped fast

Write in the exact voice you talk in, and ship it fast rather than editing it into stiffness.

"that how long does that take and then I get the sense that that"

— 03:00

The two instructions are one instruction, because editing is where the voice usually goes — each pass makes the sentence more defensible and less like anybody saying it.

04:45 - The draft runs in two passes

Draft in two passes: dump the structure as fast as possible in bullet points, then do a single narrative pass.

"basically it's a two-part process. One is an I just like try to basically"

— 04:45

The two passes have completely different jobs, and neither one of them is revision. Mixing them is what makes a first draft take a week: the structure keeps being renegotiated while the sentences are being made.

05:58 - A through-line has to be simple, punchy and aggressive

A title or through-line has to be simple, punchy and aggressive to punch through a world awash in media.

"well it's got for what I you know it's got to be something very"

— 05:58

Marketing training says the phrase has to cut to get through the saturation, and a broad word carries the claim past its original subject. A three-word rallying line was chosen to be straightforward and to widen from software to industry generally.

06:44 - Input quality is what caps output quality

Input quality caps output: spend the day talking to the leading experts on their own subjects and read every spare minute.

"Yeah. So I would say a couple things. So one is um I I"

— 06:44

Output cannot exceed input, which makes the input the actual work. Both halves of the input are named, and the expensive one is access to people rather than access to books.

09:58 - The information diet gets barbelled at both ends

Barbell your information diet — read only the super-current or the timeless, and screen out everything from yesterday back to about ten years ago.

"general method is to try to barbell my information intake. So it's either stuff"

— 09:58

Everything from yesterday back to about ten years ago gets screened out — the band that feels informative and mostly is not.

10:38 - Group chats are a way at a kind of truth in an era of public lying

Group chats are constant ambient communication and a way to get at a kind of truth in an era of high public lying.

"did you come up with this idea of group chats? Like you are the"

— 10:38

In an era of high public lying, the private channel is doing work the public one no longer can.

12:07 - Many group chats rather than one, because a closed group polarises

Run many group chats with very different people, not one — because any closed group polarises toward its own extreme over time.

"group chat idea, I think that this would be, you know, if in so"

— 12:07

A like-minded group left to talk drifts more extreme, and a meta view across several chats is what lets one be reset when it locks in or goes down a rabbit hole. The drift of a casually-inclined group into a hardened position after a year of talking only among itself is the studied case — and every person added changes the dynamic, so one wrong addition can destroy it, as at a dinner party.

14:54 - Representing others calibrates the methods as well as the topics

When you represent others you calibrate not just your topics but your methods of engagement — and the essay is the safer, fuller medium.

"Well, so then I specifically just and you know this is maybe a general"

— 14:54

Representing other people calibrates not just the topics but the methods of engagement — and the essay turns out to be the safer, fuller medium once that constraint exists.

16:51 - A big idea can be a statement of the obvious

A big idea can be a statement of the obvious; its force comes from timing, not novelty.

"Well it's crazy because you you are kind of a meme lord like you"

— 16:51

The same claim reads as provocative or not depending entirely on the moment it lands in. A the world is being eaten by software thesis felt obvious to him; the aftermath of a financial crash made it sharp. His worry at the time was that it was not provocative enough.

17:48 - The famous pieces were written fast on years of thought

The famous pieces were written fast but rested on years of discussion and thought; the trigger was frustration that people did not understand something.

"That one was written out of frustration maybe as opposed to rage. Ah, okay."

— 17:48

Written fast, and resting on years of discussion and thought. The trigger was frustration that people did not understand something, which is the same emotional charge from the opening in a different register.

18:44 - The publishing venue is one that is owned rather than rented

Publish where you own the audience, not where you rent it — legacy venues will turn you away, so build the independent audience sooner.

"happening. What does it say about your approach to distribution that that was a"

— 18:43

Legacy venues will turn you away, so the independent audience gets built sooner rather than later. Owning the distribution is the difference between having an audience and borrowing one.

20:33 - Atomic units of media recur across history and never die

Atomic units of media recur across history and never die; the internet is the meta-medium that can represent every prior form, so all of them explode back into use.

"there for there are certain like just atomic units of media that just seem"

— 20:32

The internet is the meta-medium that can represent every prior form, which is why all of them explode back into use rather than being replaced.

23:31 - What gets written down is what has actually been lived through

Write down what you have actually lived through, especially where there is an oversupply of advice from people who have not.

"What was the motivation behind your personal blog? It was the same thing. I"

— 23:31

Lived consequence is credibility competitors cannot fake, and the world is awash in advice from people who never went through the thing. The blog was started largely to answer bad startup advice from people who had not made the mistakes or lived the outcomes.

24:09 - Time around lateral thinkers produces lateral thinking

Spend time around reflexive lateral thinkers and you start thinking laterally too; never take a statement at face value.

"What' you learn from having Venka around? Breaking Smart is so good. Breaking Smart."

— 24:09

The habit transfers by exposure rather than instruction: never taking a statement at face value is caught, not taught. The capability is caught rather than taught.

25:53 - Every person added to a room changes its dynamic

Every person added to a room changes its dynamic — which is why the obvious move of simply keeping brilliant advisors on hand is rarely done.

"Why wouldn't you just hire like 10 insane lateral thinkers, put them on payroll,"

— 25:53

A model could sit in on any meeting today, with transcription and speaker disambiguation solved — and essentially nobody does it.

30:09 - Software is itself a literary genre

Software is itself a literary genre — writing code is a hybrid of writing a novel and building a bridge.

"what you want to do for the Brooklyn Bridge. Exactly. Right. So, it's not"

— 30:09

It has a structural part that must work and a creative part running all the way through, which is why every piece of software ever shipped has bugs and a bridge must not. Move fast and break things is wrong for a bridge; the best programmers write code as creative expression. It is not civil engineering, and the creative element is genuinely missing there.

31:29 - The writing interface is becoming a continuous dialogue

The writing interface is becoming a continuous dialogue with a machine, and it is coming to every form of writing, not just code.

"Yes. Uh and then what what's happened is programming languages have evolved um where"

— 31:29

The old loop was write, run, fix; the new one is a draft plus a model reading and commenting, or drafting while the writer gives feedback — pair work of human and machine. A coding assistant already comments on code in real time, and the prediction is that prose and fiction writers work the same way soon, staying in charge of the whole. It has not happened for prose yet.

32:48 - Training in code teaches writing as structure

Training in code teaches you to think of writing as structure, which shows up as density and strong logical argument.

"coincidence that Silicon Valley's contribution to writing culture has been density and very good"

— 32:48

Density and strong logical argument are what it looks like on the page: the paragraph is assembled the way a function is. The transfer runs from the engineering into the prose.

34:45 - Time is not the binding constraint on writing more

Time is not the binding constraint on writing more — the charge or the idea is.

"So what's the binding constraint on you writing more? Don't tell me time because"

— 34:45

Not time — the charge or the idea is. The constraint is upstream of the schedule, which is why more hours rarely produces more writing.

40:08 - Whoever writes the thing down holds real power

Whoever writes the thing down holds real power, independent of their formal role.

"Well, by the way, one of the things I've observed, maybe this is, it's"

— 40:08

It is an observation about organisations that happens to be an argument for learning to write. Whoever writes the memo decides which options are on the table by the time anybody reads it.

41:19 - A genuinely well-written message gets read

A genuinely well-written message gets read — near-certainly — so strong writing is a standing market inefficiency for getting attention.

"I mean the other thing is say that some rando wanted to get to"

— 41:19

Strong writing is a standing market inefficiency for getting attention, which is as close to free money as any of this gets.

42:22 - Writing became so easy that we are awash in bad content

Writing has become so easy that we are awash in bad content — and by the same tools there should be an explosion of the best content ever, yet the good stuff is still hard to find.

"So, so here's here's the paradox. Here's the paradox. It's now so easy to"

— 42:22

By the same tools there should be an explosion of the best content ever written, and the good stuff is still hard to find — which is a statement about discovery rather than about supply. The bottleneck moved to discovery.

43:41 - An arbitrary early default can harden into a permanent standard

An arbitrary early default can harden into a permanent standard.

"Why are hyperlinks blue? Cuz I like blue. [laughter] Is that why? Yes. It's"

— 43:41

The lesson is about how much a small early decision costs once everything is built on top of it. Nobody chose it as a standard; it simply survived long enough to become one, which is how most conventions get made.

44:59 - Comment quality needs a consciously calibrated system

Comment quality is not luck — it needs a consciously calibrated system to evaluate and surface quality and bury the bad.

"What's your read on why there's been such stagnation in the quality of comment"

— 44:59

Left alone it does not arrive; what is needed is a system calibrated to surface quality and bury the bad — and the web's missing feature was a universal writing layer that failed on scale.

48:09 - Internet writing became the arena where politics plays out

Internet writing became the arena where politics plays out because the internet became central to life — the dog caught the bus.

"How has online writing morphed in a way that's different from what you would"

— 48:09

Once the network became where people do, read, learn and argue about everything, and politics is central to life, its content necessarily turned hyper-political. Early network groups treated raising politics as rude; a switch flipped around the mid-2010s and politics infused everything. Part of that is attributed to the political era rather than to the network alone.

50:04 - Reading happens for content, and outstanding style punches through anyway

He reads for content, not style — but outstanding style still punches through even when you are not looking for it.

"Do you ever read for style or do you only read for content? Um"

— 50:04

He reads for content rather than style, and outstanding style punches through anyway. That is the strongest available test of style: it works on somebody who is not looking for it.

51:00 - What makes style enchanting is proportion

What makes style enchanting is proportion — the deliberate right amount of time, detail, words and emotion on each thing.

"What are the things about style that are really enchanting to you? It's just"

— 51:00

Style here is structural: everything flows, each topic gets exactly its due, and the words and the emotion are exactly right — engineered rhythm rather than decoration. The felt experience of a master biographer is the example, and he frames the reading as maybe a computer-science way of seeing it.

52:04 - The strongest writers turn real history into myth

The strongest writers turn real history into myth, and immersive concrete detail creates a sense of being there.

"right? Like certain people at the top of their games, right? Like there just"

— 52:04

Deep, kaleidoscopic detail of a real time and place makes the reader feel present, and real history beats speculation because real people actually acted. A thousand-page novel recreating a city in a week of panic sits beside a series turning a century's history into myth.

1:01:22 - The published quality gets optimised rather than the schedule

Optimise the published quality, not the schedule — a piece takes the time it takes and then it is good.

"book and ship it out the door. What do you think we could learn"

— 1:01:21

A piece takes the time it takes and then it is good. Optimising the schedule instead is what produces work that shipped on time and did nothing.

1:05:56 - The actual best arguments get engaged on a hard topic

On a hard topic, engage the actual best arguments so the informed reader concludes you understand the field — even if they disagree.

"What would your writing be missing right now? Oh, so for this one, for"

— 1:05:56

The informed reader is the one to write for, even when they disagree. Disagreement is survivable and being dismissed as uninformed is not.

1:07:02 - The big-piece process runs in five stages

The big-piece process: think and read for a long time, dump all the points, organise, fill the structure, then draft point by point.

"So, walk me through the process of how you thought about writing a big"

— 1:07:02

Five stages, and only the last one looks like writing from the outside. The four invisible stages are where the piece is decided, which is why the visible one goes quickly.

1:08:15 - A manifesto is polemic rather than observation

A manifesto is polemic — written to persuade and rally true believers, not to be a neutral observer — so it needs spirit and energy.

"yeah, and then I'm trying and then, you know, it's a little it's a"

— 1:08:15

The register is deliberately not the detached thirty-thousand-foot view, and rereading old manifestos is how the right spirit gets loaded before writing one. Historical manifesto writers could not predict what their work would later enable, which is part of what the form costs.

1:09:31 - Feedback gets sought on three specific things

Feedback gets sought on unintended anger, on clarity and on whether the piece drags stylistically.

"ask for feedback on your writing what are you trying to get feedback on"

— 1:09:31

Intended anger is fine and accidental anger is the risk worth catching; clarity checks that the thing conveys at all; the style check asks whether it bogs down. The list arrives after the joke that what he mostly wants is praise.

1:10:24 - The last paragraph becomes the first

Take the last paragraph and make it the first — order for impact, hit hard up front, and keep motivating force all the way through.

"Well, it's like the old it's like the old say the old writing advice."

— 1:10:24

You cannot read your own finished piece with fresh eyes, so a reordering rule forces the strongest material forward without requiring the judgement you no longer have. It is old writing advice, applied by habit rather than decided each time.

1:11:58 - The essay is about to become a short film

The essay is about to become a short film — generative tools will soon turn an essay into an eight-minute video, and verbal and visual creators will express far more, far faster.

"there should be far more literally animated shorts. Uh, right. If you're doing a"

— 1:11:58

Verbal and visual creators will express far more, far faster, which changes what writing an essay is for.

1:13:52 - The feed gets curated at the level of the single post

Curate an information feed at the level of the single unit — follow on one post, block liberally on one post, and keep a small set of curated lists.

"How do you think about curating your information environment on Twitter?' Yeah. So, [laughter]"

— 1:13:52

The barbell runs here too — books plus the live feed — and a single post is enough signal to follow or block on, with private lists holding the more curated intake. The lists stay private after they were used against him, and people rotate out when he changes his mind.

1:14:57 - A distinctive mind is recognisable from three sentences

You can recognise a distinctive, high-quality mind from three sentences — and a good feed surfaces exactly those people.

"and and so it basically and and it's actually this actually the power of"

— 1:14:57

A good feed surfaces exactly those people. The closing caution is that a signal chosen to imply importance backfires once everybody understands it.

Personal Branding Lessons

A conversation about writing fast and publishing on your own ground. The moves below are the method.

Write in the voice you actually talk in and ship it fast

The two halves are one instruction, because editing is usually where the voice goes. Shipping before a piece gets smoothed keeps the thing that made it worth reading, and the stiffness that replaces it is almost never noticed by the person adding it. 03:00

Dump the structure in bullets, then make one narrative pass

Two passes with completely different jobs, and neither of them is revision. The first gets everything down as fast as possible; the second turns it into prose once. Combining them is what makes a draft take a week. 04:45

Barbell the information diet

Read only the very current or the timeless, and screen out everything from yesterday back to about a decade ago. The middle band is the part that feels informative while mostly being neither new enough to act on nor durable enough to keep. 09:58

Run many group chats rather than one

Any closed group polarises toward its own extreme over time, whoever is in it. The remedy is structural — several groups with very different people — rather than a matter of choosing better company for the one you have. 12:07

Publish where you own the audience

Legacy venues will turn you away eventually, which makes the independent audience something to build sooner rather than at the point of needing it. Owning the distribution is the difference between having an audience and borrowing one. 18:44

Write down what you have actually lived through

Especially where there is an oversupply of advice from people who have not. Lived material is the part nobody else can manufacture, which makes it the only genuinely defensible supply in a market where everything else is abundant. 23:31

Engage the best arguments against you

On a hard topic, take on the actual strongest counter-arguments so that an informed reader concludes you understand the field. Disagreement is survivable; being dismissed as uninformed is not, and the second is what weak engagement produces. 1:05:56

Take the last paragraph and make it the first

Order for impact rather than for logic. Hit hard up front and keep the motivating force running all the way through — which usually means the conclusion you arrived at was the opening you needed. 1:10:24

Questions

Each answer ends at the moment in the recording where it is given.

How do I write in my own voice instead of sounding stiff and formal?

The most common cause of stilted prose is switching registers the moment you start typing — writing in a voice different from the one you talk in. The fix is to aim for the page to read exactly like a conversation with you, getting what you would actually say out and resisting the urge to edit it into something more formal. It is offered as what works for the writer and the interesting people he knows, not a universal law. 03:33

What is a fast, repeatable process for drafting an essay?

Two passes. The first exists only to get the structure out — dump fifteen or twenty points you already know you want to make, as fast as you can, in bullets, then put them in a sensible order. The second is a single narrative pass that turns the skeleton into prose. The point is not to slow the point-dump by crafting sentences too early, which is what interrupts the thinking. It is bullet points, deliberately, not a novelist's detailed outline. 04:45

How should I decide what to read and who to follow so my writing gets better?

Barbell it. Take in only the very current or the timeless and screen out the stale middle, roughly everything from yesterday back about a decade. Current means live experts and a fast feed of the sharpest people working now; timeless means books, often long stretches of audio on a single subject. The logic is that output is capped by input, and this is a reading rule about where returns actually sit. Offered as a personal method, not a rule for everyone. 09:58

Should I publish on my own site or pitch it to a bigger publication?

A widely-shared piece was turned down by the major outlets and went out on an owned property instead — and the retrospective regret was not building that independent audience far sooner. A rented venue can reject you, change its rules, or bury you; an owned channel cannot, and it keeps the relationship, which is where compounding returns accrue. The soft "but this worked out" does not blunt the direction of the advice: build the owned audience earlier than feels necessary. 18:44

Why do some obvious ideas land as big and others fall flat?

Because force comes from timing, not novelty. The most-cited thesis felt so obvious to its author that the worry was it would not be provocative enough — what made it sharp was the moment it landed, in the aftermath of a crash. The same claim reads as provocative or flat depending on when it is said. An idea does not need to be new to be big; it needs to meet a moment that makes the obvious suddenly worth stating. 16:51

How do I get busy, important people to actually read what I send them?

By making the writing genuinely good, which is rarer than it sounds. A well-written, contrarian message from someone with real domain expertise is near-certain to be read, because most inboxes and feeds are full of the opposite — and that scarcity is a standing inefficiency you can use. This is how attention actually forms around a voice: the effort visible in the writing is itself the signal that it is worth the reader's time. 41:19

Why can I not seem to write more, and is it really a time problem?

Almost certainly not time. The observation that cuts through the excuse is that those pieces were written fast, so hours were never the binding constraint. The real limit is the charge or the idea — some writers produce daily, for others it pours out at once, but neither is gated by the clock. If the writing has stalled, the honest question is whether you have something you genuinely need to say, not whether you can find the afternoon. 34:45

How do I make the opening of a piece hit hard?

Take the last paragraph and make it the first. By the time a piece is finished you are too close to read it with fresh eyes, and the strongest material has usually sunk to the bottom where you finally arrived at it. Order for impact instead: move the sentence you were building toward up to the front, hit hard there, and keep a motivating force running all the way through. It is a mechanical rule you can apply alone, without a second reader. 1:10:24

What should I actually ask for when I get feedback on a draft?

Three specific things. First, whether it makes people angry for unintended reasons — intended provocation is fine, accidental offence is the risk to catch. Second, clarity: does it make sense and convey the thing. Third, whether it drags stylistically, whether it bogs down. Praise is pleasant but not information. Note that feedback calling a piece too polemic can be exactly the point when you are writing for effect rather than to be a detached, neutral observer. 1:09:31

Sources

Marc Andreessen: Blogposts, Tweetstorms, and Manifestos

The sections above follow the episode's own order. Each timestamp is the point in the recording where the idea is discussed, so the post reads as a map of the conversation rather than a rearrangement of it.

Quotations come from the episode transcript with spoken filler ("okay", "right", "you know", "um", "actually", "kind of", "sort of") and false starts removed. No words are added: square brackets mark anything inserted for sense, and an ellipsis marks any cut. British spellings are restored, and speech-to-text errors were repaired against context only where the intended word is unambiguous. Where a line appears both in a cold-open teaser and in its place in the conversation, the timestamp cites the second — the point at which it is actually said.

Quotes are stamped with the moment they are said rather than with a speaker's name.