Think in public like Andreessen: a field guide

Develop a thesis, not an opinion

Make a claim about the world. An opinion tells people how you feel about an event; a thesis explains what happens next, why you believe it, and which facts would prove the argument wrong. A claim is something you can stand by; an opinion is a vague musing. You want the claim, if you want to make anyone sit up and pay attention.
Write a sentence people can test. Readers can measure "Software is eating the world" against revenue, employment, market value, and industry structure. They can't do much with "Technology will become more important." Testable writing = bold, dangerous and risky. You're playing with the very real possibility of being very publicly wrong. This is not a bad thing.
Look further ahead. Plenty of writers can explain this morning's announcement. The more useful work is connecting today's evidence to the conditions that may // may not exist ten years from now. It's the difference between breathless idiots posting "BREAKING" on Twitter and someone who actually has a long view.
Give each essay one job. In his strongest essays, Andreessen makes one consequential claim and then follows it through industries, incentives, objections, and likely outcomes. Have a point, and make that point singular.
Tell people how to act. Give the founder something to build, the investor something to fund, or the worker something to learn. If the thesis doesn't shape a single decision, if you can't extrapolate direct action from it, it's weak. Keep it in your drafts.
Pick an argument worth having. A serious thesis contradicts someone's model, threatens someone's investment, or challenges someone's status. A blog post that does none of those things is dead air. Unless someone can disagree with you, you might as stay silent. The effect is broadly the same.

Publish for the archive

Write the definitive version. When people ask what you think about a subject, give them one complete answer rather than a trail of disconnected posts.
Publish somewhere durable. Social platforms can circulate the idea, but the full argument should live somewhere people can search, link, and retrieve years later. See also: Paul Graham's blog.
Keep the original date visible. Readers should know whether you made the call before the market agreed, during the transition, or after the conclusion had become obvious.
Keep the misses too. Preserve the original claim, add a correction, and explain the evidence that changed your mind. A spotless archive can only belong to a liar who refuses to admit their mistakes.
Remove today's headline. Ask whether the argument would still help someone five years from now. If it collapses without the news hook, it is probably commentary rather than a durable thesis. Leave the commentary to the pundit class. Think better // think higher.

Put real stakes behind the argument

Identify the consensus. State the dominant position in terms its supporters would accept, then show which assumptions you think fail.
Replace hedges with conditions. Instead of saying "this might happen," explain which conditions would produce the outcome and which facts would rule it out. Be definite. Be sure. Be concrete.
Use numbers when you have them. Cost curves, adoption rates, margins, market share, and deadlines make claims easier for other people to test. If they're easier to test, it's easy to be proven wrong. But it's equally easy to prove you were right and do it in a way people remember.
Follow the claim into actual decisions. Explain which companies gain power, which institutions lose it, which jobs disappear, and which opportunities open if your thesis is right.
Say how you would know you were wrong. Identify the facts that would weaken or kill the thesis before a critic has to do it for you.
Answer the best objection. Present the opposing case as its strongest advocate would. If your argument only survives against a caricature, the argument needs more work. You can hide behind a flimsy contra, but you're only hiding from yourself. And sooner or later, someone's going to point out that the emperor is a coward with no conviction.
Disclose your interest. Tell people when you own the asset, fund the sector, sell the product, or benefit when others accept the idea. Strive for maximum transparency. It'll make every argument stronger.

Create language that people can reuse

Find the shortest honest version. "Software is eating the world" compresses an argument about companies, industries, labor, and capital into six words.
Invent fewer terms. Create a term only when ordinary language cannot express a distinction your argument needs.
Test the line aloud. Someone should be able to repeat the central claim after one reading without turning it into mush.
Prefer ordinary words. Technical vocabulary helps specialists communicate, but decorative jargon always hides weak reasoning and narrows your audience. If a 5th grader can't understand your blog post, congratulations - you're smarter than me. But you're also unlikely to garner any attention outside the competing categories of sycophants who agree with you, and Linux aficionados who don't.
Make the terms fit together. One phrase can attract attention, but several connected concepts give your readers a durable way to interpret companies, markets, and institutions.
Never guard the phrase. People may use it without credit. If the idea travels widely, that reach will usually matter more than constant attribution.

Publish near the edge of your certainty

Publish while the outcome remains open. Once the evidence has settled every serious objection, the thesis carries less risk and offers people less they do not already know.
Separate fact, inference, and forecast. Tell readers which claims the evidence establishes, which conclusions you have drawn, and which events you expect next.
Say how sure you are. Use "I know," "I believe," and "I suspect" according to the evidence behind each claim. Do not hide uncertainty behind a confident tone.
Keep the live question visible. Show the unresolved tension, the competing explanations, and the evidence you still need.
Do not publish a pile of fragments. You do not need every answer, but you do need verified facts, defined terms, a causal account, and a serious response to the best objection.
Check every factual claim. An uncertain forecast can still rest on solid evidence. Predictive risk does not excuse careless sourcing.
Change the theory when the facts contradict it. Do not protect a favored framework by explaining away each failed prediction.
Correct errors in the same place. Quote the original claim, identify the mistake, explain the new evidence, and state your revised position.

Turn publication into a discipline

Publish enough to develop judgment. Repetition shows which arguments survive criticism, which phrases travel, and which assumptions collapse.
Let publication expose the gaps. Private notes allow you to skip steps. Public claims force you to define terms, support causal links, and answer people who disagree.
Judge the full body of work. No single essay has to carry your reputation. Over time, people will judge the record across recurring questions, correct calls, failed predictions, and visible revisions.
Choose a schedule you can keep. Weekly posts mean little if you stop after two months. Publish at a pace you can sustain for years.
Use the right length for the job. Test a claim in a short post, make the full case in an essay, and use a series when one piece cannot hold the whole argument.
Return with new evidence. Revisit old questions when you can add facts, correct assumptions, or trace consequences you could not see earlier.
Pay attention to behavior, not applause. Likes record a momentary reaction. Serious rebuttals, unsolicited citations, changed decisions, and references years later tell you more.

Use criticism as evidence

Sort each objection before answering it. A critic may dispute your facts, logic, forecast, values, or incentives. Each dispute calls for a different response.
Do not let popularity settle the claim. Large audiences reward confidence, novelty, and conflict, but none of those qualities makes an argument true.
Keep smart dissenters close. Find people who understand the subject, oppose your conclusion, and can identify the weakest point without distorting your case.
Record each change of mind. Write down the facts, arguments, and experiences that forced the revision. Over time, the record will expose your recurring errors.
Check what happened after people acted. When founders, investors, or policymakers use your thesis, study the outcome and revise your position when the evidence calls for it.
Connect the original claim to every update. People should be able to follow both the conclusion and the reasoning that produced it.

Become a reference point

Give people a way to think. Commentary joins an existing conversation, while a durable framework helps people make decisions after the original event has passed.
Write for decisions. Help someone choose what to build, fund, study, hire for, or stop doing.
Keep the idea useful after the headline fades. A strong explanation can keep helping people across companies, industries, and market cycles.
Expect many pieces to disappear. Few essays become standard references. The others still help you develop the judgment and language needed to write the rare ones that do.
Make the call while it can still help. One clear thesis, published before consensus forms, can shape decisions you may never hear about.

Follow this discipline for years and people may start using your essays to orient themselves. They may quote your phrases, apply your frameworks, and send old pieces to people who have never encountered your current work.

Andreessen built influence by making consequential claims before anyone knew the outcome. He attached evidence, answered critics, and preserved enough of the record for people to judge the work over time.

The lesson is not to sound certain for its own sake. It is to state the thesis early, show your reasoning, correct the record when you are wrong, and keep writing.