Tech Twitter: current, curated evidence from technology discourse
Tech Twitter is a read-only research surface 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, filters the resulting posts for relevance, and publishes only material that has completed its publication workflow. Use it when you need current evidence about artificial intelligence, developer tools, startups, product development, company building, or shifts in the language and priorities of the technology industry. The homepage is the publicMonitor database workspace: its panels summarize the latest posts, historical comparisons, trending items, active topics, notable authors, and discussions with unusually high participation. The News page turns that daily signal into a concise email brief.
When an agent should use Tech Twitter
Reach for Tech Twitter when a task depends on what practitioners are saying now rather than on timeless background knowledge. Good jobs include finding the launches that matter today, explaining what changed in a topic during the last day or week, locating primary posts behind an emerging narrative, comparing competing arguments, assembling a short reading list, or checking which people are driving a discussion. Use Trending Now for the current story board; that is a REST discovery skill, not a context-bundle kind. The Time Travel pages are useful when the date itself matters: they provide shareable daily views instead of forcing an agent to reconstruct a past feed from present-day rankings. Topic and profile pages are better when the question is about a subject or author across a longer period. Articles and Threads provide longer editorial context, while canonical tweet pages preserve the individual evidence item and its source link.
How to use the evidence responsibly
Treat each result as evidence of a public statement, not as proof that the statement is correct or that it represents consensus. Prefer several independent posts when making a broad claim. Keep author, timestamp, topic, engagement, and quality information attached to the text that they describe. Cite the canonical Tech Twitter URL supplied with an item so another reader can inspect the same record, and follow the original source URL when exact wording or media matters. Rankings are discovery aids: engagement can reveal attention or disagreement, but it does not establish truth, importance, or endorsement. Summaries should distinguish a product announcement from commentary, an argument from an agreed conclusion, and a recent signal from a durable trend.
Evidence and discovery for agents
Start with the Evidence for Agents page for citation-ready evidence, context, and receipts from published records. It explains how to cite a canonical URL and what a receipt is not: official X access, a firehose, or a way to act on an account. The concise llms.txt guide explains when to use Tech Twitter and points to the most important resources; the expanded llms-full.txt guide includes evidence recipes and longer worked examples. The agent card is machine discovery metadata for the read-only evidence agent. Use the sitemap index to discover indexable topics, profiles, articles, tweet records, trend days, and quality-filtered historical days. The machine contract remains available for clients that already know how to consume it; the Evidence for Agents page does not document that contract.
Choose the smallest useful reader surface
The live Trending view is the quickest place to check which published posts are receiving attention now. Use Time Travel when a question names a date or asks what people knew and discussed at a particular moment; its daily URLs remain shareable after the live monitor moves on. Browse the Topics index for a subject-level history and the Profiles index for an author-level history. Choose Articles when the source is a long-form X article and Threads when an editorial recap is useful for orientation. The Skills directory is a separate catalog of reusable agent instructions; it is not the evidence index itself. Starting with the narrowest surface reduces duplicate results and makes it clearer whether a date, topic, author, ranking signal, or editorial summary determined what was returned.
Understand the evidence record
Public evidence rows identify their source type and preserve a stable record ID, title or source text, publication timestamp, author when available, topic references, engagement metrics, quality score, and a short explanation of why the item was included. Provenance fields name the underlying public record, while canonicalUrl points to the Tech Twitter page suitable for a citation and sourceUrlpoints to the original X material when it is available. A summary is a navigational aid rather than a quotation. If exact language changes the answer, inspect the original source before quoting it. Null fields should remain null instead of being guessed, and engagement counts should be described as observed metrics rather than converted into claims about credibility or agreement.
Freshness, limits, and recoverable errors
Live rankings and monitor panels can change as published material and metrics change; permanent item and date URLs are the better citations. Context requests accept bounded windows of 24 hours, 48 hours, or seven days and return a limited evidence set so an agent can inspect every row it uses. Public budgets protect the shared service. A rate-limit response includes retry guidance, so wait rather than sending parallel retries. REST failures use JSON error objects with a stable code, readable message, and resolution hint. A2A calls retain JSON-RPC errors. Unknown API paths return JSON, while unknown reader paths return a real 404 with links to the OpenAPI document, agent instructions, sitemap, agents page, and homepage. Those recovery links are preferable to guessing another URL.
A practical evidence workflow
First restate the question with an explicit topic and time window. Usekind=auto only when the question itself clearly signals a launch, argument, reading request, narrative, or recent change; otherwise select the recipe directly. Review the returned instructions and facets, then inspect the highest-ranked evidence rows and follow related endpoints only when the first bundle leaves a material gap. For a broad conclusion, look for more than one author and separate original reporting or product announcements from reactions. Record the generated time and window beside volatile metrics. In the final answer, cite canonical item URLs next to the claims they support, name uncertainty or missing coverage, and avoid presenting a curated sample as a complete account of X or of the technology industry.
Local and offline agent workflows
Stateless HTTP requests are best for a fresh, bounded lookup. Agents that need a local SQLite mirror, offline full-text search, cited 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. Follow the linked skill's current install and verification instructions before invoking it. The integration is distributed outside techtwitter.com, cannot write to X, and does not expand the curated corpus into an arbitrary X search. Use live Tech Twitter evidence when current upstream state matters; use the local mirror when repeatability, offline work, or change since a previous sync matters more.
What the public service does not do
The public interfaces are designed for bounded retrieval and discovery. They do not provide a user account, a browser-session export, a way to post to X, or a fallback scraper for private or unavailable material. Public API routes may enforce durable rate limits and return only published records. A missing item can mean that it was never monitored, did not pass curation, has not completed processing, or is no longer available; absence should not be interpreted as evidence that an event did not happen. For broad exploration, begin with the sitemap or topic pages. For a time-sensitive answer, begin with agent context or the latest monitor data, narrow by topic and time window, then cite the returned canonical evidence rather than relying on a Monitor panel label alone.
Trust, ownership, and contact
Read About Tech Twitter for the service's purpose, sourcing, and editorial boundaries. Use the contact page for corrections, developer questions, newsletter help, or security-reporting guidance. The privacy notice explains newsletter email handling, local browser state, optional analytics, and the durable controls used to protect public endpoints.
@claudeai · 16hWe’re opening a research preview of Claude Duet, a new app where Claude works as a proactive agent.
It comes to you with what needs your attention, completes tasks you ask of it, keeps working when you step away, and checks with you before it acts on your behalf.
@eastdakota · 10hSome stats on #php2shell. In the 10 hours before it was publicly announced, across the @Cloudflare network we saw thousands of attempts to exploit it. That jumped to tens of thousands in the 45 minutes post announcement, hundreds of thousands within 2 hours, and millions now.
@thsottiaux · 5hDay 17/ Bet you thought we forgot, but no, it’s reset day today.
Also good time to remind you to get Daybreak Blue access, it only takes 15 mins and you don’t want to have to do it in the middle of an incident.
@openai · 21hOpenAI models helped researchers—operating as part of our Patch the Planet initiative—discover critical security flaws in PHP, an open-source programming language used to run an estimated 70% of websites.
Over the past two months, we led a coalition of maintainers, researchers, and security providers to prepare fixes and protections ahead of disclosure. This duration was necessary because the fix was quite intricate and required a major change to the PHP engine. Every site using PHP can now safely upgrade their version to stay secure, in addition to applying mitigations from various security vendors.
Due to the severity of these flaws and PHP’s widespread use, we consider this bug to be one of the most impactful vulnerabilities in recent memory. It also marks a new phase, where AI can find security bugs with internet-scale impact.
This is an example of the Defender’s Window in action: to our knowledge, frontier AI was able to find this vulnerability ahead of attackers, giving defenders time to prepare.
https://openai.com/index/coordinated-disclosure-of-critical-vulnerabilities/
Claude
Opus 5.5
ChatGPT
Codex
Muse
Grok Bot
Jev
