Tech Twitter is a read-only research interface for understanding what technology builders, founders, investors, and researchers are discussing on X. The site monitors a curated set of public accounts through the official X API, applies relevance filtering to the resulting posts, and publishes only content that has completed the publishing process. It is aimed at scenarios that need current evidence about artificial intelligence, developer tools, startups, product development, company building, or changes in the language and priorities of the technology industry.
Uses Use Tech Twitter when what you need is what practitioners are saying right now, rather than timeless background knowledge. Suitable tasks include: discovering launches that matter today, explaining how a topic has changed over the past day or week, locating the original posts behind an emerging narrative, comparing competing viewpoints, compiling a short reading list, or seeing who is driving a discussion. Use Trending Now to view the current storyboard; that is a REST discovery skill, not a context bundle type. When the date itself matters, the Time Travel pages are useful: they provide shareable daily views without requiring an agent to reconstruct a past information flow from current rankings. When the question concerns a topic or author over a longer period, topic pages and author pages are more suitable. Articles and Threads provide longer editorial context, while canonical tweet pages preserve a single item of evidence and its source link.
Using evidence responsibly Treat each result as evidence of a public statement, not as proof that the statement is correct or represents consensus. When making broad claims, prefer multiple independent posts. Keep the author, timestamp, topic, engagement, and quality information together with the text they describe. Cite the canonical Tech Twitter URL provided with the item so that other readers can inspect the same record; when the exact wording or media matters, follow the original source URL. Ranking is a discovery aid: engagement can reveal attention or disagreement, but it does not establish truth, importance, or endorsement. Summaries should distinguish product announcements from commentary, arguments from agreed conclusions, and recent signals from durable trends.
Evidence and discovery for agents Start with the Evidence for Agents page to get citable evidence, context, and credentials for published records. It explains how to cite canonical URLs and what the credentials are not: not official X access, not a firehose, and not a way to take action on accounts. The concise llms.txt guide explains when to use Tech Twitter and points to the mostImportant resources; the extended llms-full.txt guide contains evidence recipes and longer examples. The agent card is machine discovery metadata for the read-only evidence agent. Use the sitemap index to discover indexable topics, author pages, articles, tweet records, trend days, and quality-filtered historical days. The machine contract remains available to clients that already know how to consume it; the Evidence for Agents page does not document that contract.
Choose the smallest usable reading surface. The real-time Trending view is the fastest place to check which published posts are currently receiving attention. When a question specifies a date or asks what people knew and discussed at a particular moment, use Time Travel; its real-time monitoring continues, and daily URLs remain shareable. Browse the Topics index for topic-level history and the Profiles index for author-level history. Choose Articles when the source is a long-form X article, and Threads when editorial review helps with locating. The Skills directory is a separate directory of reusable agent instructions; it is not the evidence index itself. Starting from the narrowest surface reduces duplicate results and more clearly shows what determines the returned content: date, topic, author, ranking signal, or editorial summary.
Understand the evidence records. A public evidence row identifies its source type and preserves a stable record ID, title or source text, publication timestamp, author (if any), topic references, engagement metrics, quality score, and a brief explanation of why the item was included. The source field names the underlying public record, while canonicalUrl points to the Tech Twitter page suitable for citation and sourceUrl points to the original X material (if any). The summary is a navigation aid, not a quotation. If the exact language would change the answer, check the original source before citing. Empty fields should remain empty rather than being guessed; engagement counts should be described as observed metrics rather than converted into claims about credibility or consistency.
Freshness, limitations, and recoverable errors. Real-time rankings and monitoring panels change as published material and metrics change; permanent item URLs and date URLs are better citations. Context requests accept bounded windows of 24 hours, 48 hours, or seven days and return a limited evidence set so that the agent can inspect every row it uses. Public budgets protect the shared service. Rate limit responses include retry guidance, so wait rather than sending parallel retries. REST failures use a JSON error object with a stable code, a readableMessages and resolution hints. A2A calls preserve JSON-RPC errors. Unknown API paths return JSON, while unknown reading paths return a real 404 and link to the OpenAPI documentation, agent instructions, sitemap, agents page, and homepage. These recovery links are preferable to guessing another URL.
Practical evidence workflow First restate the question with a clear topic and time window. Use kind=auto only when the question itself clearly indicates a release, debate, reading request, narrative, or recent change; otherwise choose a recipe directly. Review the returned instructions and facets, then inspect the highest-ranked evidence rows, and follow up with related endpoints only when the first bundle leaves a substantial gap. For broad conclusions, look for more than one author and distinguish original reporting or product announcements from reactions. Record the generation time and window alongside volatile metrics. In the final answer, place the canonical item URL next to the claim it supports, state uncertainty or missing coverage, and avoid presenting a curated sample as a complete account of X or the tech industry.
Local and offline agent workflow Stateless HTTP requests are best for fresh, bounded queries. Agents that need a local SQLite mirror, offline full-text search, citation bundles without another network call, or comparisons across repeated syncs can consult the Printing Press Tech Twitter skill. Its external read-only CLI adds workflows such as since, momentum, narrative, digest, and offline evidence. Before calling it, follow the linked skill's current installation and verification instructions. This integration is distributed outside techtwitter.com, cannot write to X, and does not extend the curated corpus into arbitrary X searches. Use live Tech Twitter evidence when the current upstream state matters; use the local mirror when reproducibility, offline work, or changes since the last sync matter more.
What the public service does not do The public interface is designed for bounded retrieval and discovery. It does not provide user accounts, browser session exports, a way to post to X, or a fallback scraper for private or unavailable material. Public API routes may enforce persistent rate limits and return only published records. A missing item may mean it was never monitored, did not pass filtering, has not yet finished processing, or is no longer available; missing should not be interpreted as evidence that an event did not occur. For broad exploration, start from the sitemap or topic pages. For time-sensitive answers, start from the intelcontext or the latest monitoring data, narrow the scope by topic and time window, and then cite the canonical evidence returned, rather than relying only on Monitor panel labels.
Trust, Ownership, and Contact Read About Tech Twitter to understand the service's purpose, sources, and editorial boundaries. Use the contact page for corrections, developer questions, newsletter help, or security reporting guidance. The privacy statement explains newsletter email processing, local browser state, optional analytics, and persistent controls used to protect public endpoints.
The official text also contains several example tweets and items, such as: mcp-extensions just passed 1k GitHub stars and released v0.2, supporting editing drafts in a new conversation or an existing conversation, etc.; a user said they do not like Antigravity and asked where Google's super app is; an opinion holds that smart speakers were ten years too early, Amazon once had 10,000 people working on Alexa/Echo and lost $25B; a post was surprised that two new frontier models appeared on two consecutive days; a post said Google's new frontier model Gemini 4 Argon beats GPT-6 Astra, Claude Fable 5.1, and Opus 5.5 on most benchmarks; a discussion asked whether the world's largest enterprises are afraid to use Chinese models; a post mentioned that Apple added a new "Daily Cash" app icon in macOS 27.2 beta 4; a post said the Google Cloud Run Hackathon lands on Product Hunt tomorrow, deploy on Cloud Run and publish on Wednesday, and the winner gets $500 in @googlecloud credits; a post said Gemini 4 Argon is now available in Cursor, scoring 52.8% on CursorBench at default effort, comparable to Opus 5.5, with pricing at $2/M input; a post mentioned that when founders speak at YC they always ask: in the AI era, how much standard startup advice still applies. The homepage also lists popular items, such as Google launches Gemini 4 Argon across global apps, API, and Search (19 posts, first appeared 4 hours ago), X bundles Cursor, Grok Bot, and Premium into one subscription (7 posts, first appeared 21 hours agoappeared), Anthropic rebuilds source control for agent-scale development (3 posts, first appeared 4 hours ago), Anthropic hardens Claude against model distillation attacks (1 post, first appeared 20 hours ago), OpenAI model finds critical Chrome vulnerability in six days (1 post, first appeared 2 hours ago), Ithaca Holdings acquires Bumble for $1.3 billion (2 posts, first appeared 5 hours ago), White House moves to directly control AI model testing (1 post, first appeared 6 hours ago), Microsoft adds Anthropic models to Copilot (1 post, first appeared 23 hours ago).
Limitations and additional notes: The above tweet examples and popular items are presented only as evidence of public statements and do not prove their correctness or represent consensus. Tech Twitter's public interface does not provide user accounts, posting functionality, or scraping of private materials. Regarding specific fees, prices, and platform availability, the information provided in the official text is limited, and some content is not yet available.