GitHub Outage Map

Community Discussion

Tips? Frustrations? Share them here. Useful comments include a description of the problem, city and postal code.

Beware of "support numbers" or "recovery" accounts that might be posted below. Make sure to report and downvote those comments. Avoid posting your personal information.

GitHub Issues Reports

Latest outage, problems and issue reports in social media:

• 
  Nitin Pulyani (@npulyani) reported 38 seconds ago

  I hosted my site on github pages. Now I need an edge compute layer and a server side deployment. Migrating to vercel. Any better suggestion?

• 
  Harshil Tomar (@Hartdrawss) reported 4 minutes ago

  you should 100% be claudemaxxing i've tried every plugin, wrapper, and github repo promising to give my agent superpowers cursor rules generators, custom mcp stacks, orchestration layers that do things without even telling you whats happening under the hood honestly? i cant tell if any of them moved the needle reason is simple. you get better at something the more you use it. every abstraction layer you add is a thing you stop understanding AI labs are shipping at insane speed. if you're missing a feature, wait 3 weeks. it'll ship. so here's my actual claude setup. no plugins. no wrappers: 1/ cursor for file-level edits. claude code for full feature builds. they do different things, stop using them the same way 2/ supabase MCP connected directly. it reads and writes the db without you copy-pasting schema into every prompt 3/ chrome devtools MCP connected. claude can inspect the dom, read console errors, and debug frontend issues without you relaying anything 4/ write a CLAUDE[.]md file at the root of every project. it loads context automatically every session. no re-explaining your stack, conventions, or rules ever again 5/ ask claude code to write your cursor rules file. let the AI configure the AI. sounds stupid, works every time 6/ use sub-agents for long tasks. one agent plans, another executes. context stays clean and outputs get sharper end of day: ask it to write the *** commit message and update the docs. 10 minutes of work you will never do manually again been averaging 10-20% faster task completion since i stopped trying to hack the tools and just learned them Now go claudemaxx !

• 
  Ivail (@Ivaiill) reported 11 minutes ago

  @Shah_45_ @smasithick As far as I know the Devs are working on the problem - I read on their github page.

• 
  Alessio Persico (@alepbuilds) reported 12 minutes ago

  @Zenysi_ Happy to connect, current building a devtool for solving my own problem in searching validation around hackernews GitHub and more. What are currently working on?

• 
  Alpha Signals on X (@Alpha_Signal_X) reported 12 minutes ago

  @AnthropicAI @claudeai @DarioAmodei One week NO response from Claude Support on paid account that needs backend fix. How can this company IPO at 1T when it can even respond to emails, GitHub or Fin AI? This market is too competitive to be loyal to un loyal companies.

• 
  Eji (@mildsky1215) reported 12 minutes ago

  @ntotao @xai Cleanest setup: ~20 real GitHub issues that ship with test suites, run each model agentically until tests pass (cap at N turns), then log median turns-to-green, total $, and % solved. SWE-bench Verified is the closest public proxy but it under-weights turn-count and cost, and turn-efficiency is the number nobody publishes.

• 
  gitbankbot (@gitbankbot) reported 12 minutes ago

  open source contributors solve real problems. they should get paid when those solutions land, not after three follow-up messages. gitbank automates the full bounty flow on Base L2. maintainer sets a USDC amount on a GitHub issue. contributor merges a qualifying PR. smart contract releases the funds automatically. the payment is as reliable as the CI pipeline.

• 
  Aakash Gupta (@aakashgupta) reported 15 minutes ago

  The PM role just split in two. One group ships daily. The other ships quarterly. Same title, completely different job. Arize's CPO showed the dividing line live. She opened Claude Code with an empty directory and in under 45 minutes had a PM agent pulling 40 GitHub discussions, 60 issues, and 8 releases, scoring every single one by priority, evaluating its own accuracy, and feeding corrections back into itself on a cron. Four asks in a terminal. Build the agent. Instrument it. Suggest an eval. Run the loop. No IDE. No engineering partner. The agent caught bugs being systematically underscored on its first pass. Feature requests were getting higher priority than production issues. That kind of scoring drift normally takes a PM weeks to notice through manual backlog review. The agent surfaced it before she finished explaining what a trace is. She confirmed same-day shipping at her company. Issue arrives, PM spots it through the taste agent, Claude Code prototypes a fix, ships that afternoon. Arize has raised $131M and is hiring more *** than ever. The *** they hire look indistinguishable from engineers. She asked Claude to generate the eval knowing it would be noisy. Then used her own PM judgment to tell it where the scoring was wrong. Her taste refined the eval. The refined eval improved the agent. The improved agent generated better traces. Better traces sharpened the next eval. That loop runs overnight while she sleeps. The PM's alpha used to be consuming more user feedback than anyone else on the team. The agent consumes all of it now. What remains is the eval: defining what "good" looks like. Curating taste at scale. Any PM running traces and evals on their agents is already top 1%. The other 99% are still manually processing backlogs that an agent can score, evaluate, and self-correct every hour. By next year, that gap shows up in job descriptions.

• 
  Adam (@adam_narozniak) reported 18 minutes ago

  @OpenAIDevs Quite useless since you still allow access to only 50 most recent chats. Older ones dissapear from the sidebar and you can't open them even if they are found using this new search :( there're quite a few issues on Github reporting it

• 
  Donald Lekgwathi (@DonaldLekgwathi) reported 29 minutes ago

  @antirez That's a huge generalisation. While I do agree that Github issues is not the correct place to discuss it, they do have a valid point.

• 
  Kashan Ahmad (@thekayshawn) reported 32 minutes ago

  @denbvk @didericis @ThePrimeagen I can assure you most developers like to write code in an editor instead of reviewing it on GitHub, you're the exception in that. Infact, the whole problem with agentic coding is that developers feel distant from code which why editors won't go anywhere for a long time.

• 
  AGBABIAKA (@roidesdieux) reported 32 minutes ago

  @github fix your mess!

• 
  Josip Herceg (@JosipHercegg) reported 32 minutes ago

  @OdedSharir Not using GitHub Issues but I gave you a follow. :)

• 
  Toyesh Chakravorty (@Bhushindo) reported 32 minutes ago

  The idea started with a problem I had actually faced myself. After graduating, I lost access to my university credentials. Course notes, assignments, everything gone. The only thing left were the project works I had pushed to GitHub. That became the starting point.

• 
  vish (@vishctx) reported 36 minutes ago

  @OfirPress There’s a massive blind spot in the benchmarks. By the time an issue makes it to GitHub with a reproducible state, 80% of the hardest engineering work is already done. Current benchmarks hand models extremely precise problem statements. But in the real world, like when debugging the Linux kernel, you rarely start knowing what the problem actually is. All a user will report is “the app is OOMing, and increasing memory doesn’t help.” Digging into that requires intuition built from past issues. The root cause could be memory leaks, memory fragmentation, or a race condition where threads acquire memory and never release it leading to starvation. We desperately need benchmarks with highly ambiguous starting conditions to test if a model can navigate a state with multiple distinct root-cause scenarios. Right now, models like Opus easily get stuck in loops during open ended investigations. They rarely move forward unless I ask it to check for hypotheses A, B, or C. The next frontier for SWE evals should also include cases where the model is trying to figure out what's actually broken in the first place.