June 18, 2026

# Nightly Librarian — Newsletter draft

Run: 69eb65fb-6942-4591-8e2a-6e310f6e1b7d
Started: 2026-06-18T11:03:04.822Z
Completed: 2026-06-18T11:12:59.995Z

## Worth attention

- **I Could've Rickrolled the Entire FIFA World Cup. All I Needed Was My ID**
  https://bobdahacker.com/blog/fifa-hack
  Researcher says FIFA exposed live stream controls and match systems behind client-side auth checks during the World Cup.
- **GLM-5.2 is the new leading open weights model on Artificial Analysis**
  https://artificialanalysis.ai/articles/glm-5-2-is-the-new-leading-open-weights-model-on-the-artificial-analysis-intelligence-index
  Artificial Analysis puts GLM-5.2 at the top of open weights, with better agentic scores but heavier token use.
- **Making GHC upgrades easy**
  https://blog.haskell.org/making-ghc-upgrades-easy/
  The GHC team says newer compiler releases should become much easier to adopt by loosening base/GHC coupling.
- **v0.30.10**
  https://github.com/ollama/ollama/releases/tag/v0.30.10
  Ollama 0.30.10 adds MLX support for Command A and North models on Apple Silicon and bumps llama.cpp.
- **Elevated errors for Claude Opus 4.8**
  https://status.claude.com/incidents/k8563n9wd0r8
  Anthropic resolved a Claude Opus 4.8 error spike after affecting chat, API, and coding surfaces.

## Full digest

- [R] [vercel-changelog] New build and deploy capabilities in Vercel CLI — https://vercel.com/changelog/new-build-and-deploy-capabilities-in-vercel-cli — Resurfaced Vercel CLI changelog from July 2022, not new.
- [R] [vercel-changelog] Hydrogen projects can now be deployed with zero configuration — https://vercel.com/changelog/hydrogen-projects-can-now-be-deployed-with-zero-configuration — Old Vercel Hydrogen deploy note from July 2022 resurfaced.
- [R] [vercel-changelog] Vercel Edge Functions are now in public beta — https://vercel.com/changelog/vercel-edge-functions-are-now-in-public-beta — Old Vercel Edge Functions beta post from June 2022 resurfaced.
- [R] [vercel-changelog] IP Geolocation now available for all plans — https://vercel.com/changelog/ip-geolocation-now-available-for-all-plans — Old IP geolocation availability note from 2021 resurfaced.
- [R] [vercel-changelog] Edge Functions are now available in Public Beta — https://vercel.com/changelog/edge-functions-are-now-available-in-public-beta — Another stale Vercel Edge Functions beta changelog from 2021.
- [R] [reddit-localllama] GLM-5.2 (max) is currently the third best model available, across both open and proprietary. — https://www.reddit.com/r/LocalLLaMA/comments/1u832oh/glm52_max_is_currently_the_third_best_model/ — Community claim about GLM-5.2 Max ranking, but better covered by stronger sources.
- [R] [reddit-localllama] It looks like Rio 3.5 397B could've simply been a semi-failed embezzling of funding — https://www.reddit.com/r/LocalLLaMA/comments/1u84f4j/it_looks_like_rio_35_397b_couldve_simply_been_a/ — Speculative funding scandal thread with no verified reporting in the queue.
- [R] [reddit-localllama] GLM-5.2 is the first open-weights model to cross 80% on Terminal-Bench and beats every other open model available — https://www.reddit.com/r/LocalLLaMA/comments/1u7mexd/glm52_is_the_first_openweights_model_to_cross_80/ — Another GLM-5.2 benchmark claim, duplicated by better reporting.
- [R] [reddit-localllama] GLM-5.2 is now 1st on Design Arena — ahead of the now unavailable Claude Fable 5. — https://www.reddit.com/r/LocalLLaMA/comments/1u7qti8/glm52_is_now_1st_on_design_arena_ahead_of_the_now/ — Design Arena ranking post for GLM-5.2, but still duplicate launch chatter.
- [R] [reddit-localllama] zai-org/GLM-5.2 is here! — https://www.reddit.com/r/LocalLLaMA/comments/1u7kcwf/zaiorgglm52_is_here/ — Launch thread announcing GLM-5.2 open weights, superseded by better evidence.
- [R] [reddit-localllama] Hashicorp founder thinks local models "aren't good ENOUGH yet" — https://www.reddit.com/r/LocalLLaMA/comments/1u7nph8/hashicorp_founder_thinks_local_models_arent_good/ — Opinion thread about local model quality, not a concrete change or report.
- [R] [reddit-localllama] GLM-5.2: Built for Long-Horizon Tasks — https://www.reddit.com/r/LocalLLaMA/comments/1u84zan/glm52_built_for_longhorizon_tasks/ — Another GLM-5.2 performance thread without incremental signal.
- [R] [reddit-localllama] Is Le Gros Chaton opensource? — https://www.reddit.com/r/LocalLLaMA/comments/1u7pj4i/is_le_gros_chaton_opensource/ — Question thread about an open-source cat model, not a decision-changing update.
- [R] [reddit-localllama] Mistral - New family of open-weight models @ July — https://www.reddit.com/r/LocalLLaMA/comments/1u7klvv/mistral_new_family_of_openweight_models_july/ — Unsourced teaser about future Mistral models with no concrete release.
- [R] [reddit-localllama] Local models went from mostly useless to actually useful really fast. What changed? — https://www.reddit.com/r/LocalLLaMA/comments/1u85t9c/local_models_went_from_mostly_useless_to_actually/ — General discussion on why local models improved, not a discrete event.
- [R] [reddit-localllama] VibeThinker-3B: what is this witchcraft? Killing it at MathQA like it has ~30B parameters — https://www.reddit.com/r/LocalLLaMA/comments/1u7u3ls/vibethinker3b_what_is_this_witchcraft_killing_it/ — Single benchmark excitement post for VibeThinker-3B, lacking broader confirmation.
- [R] [reddit-localllama] Cheapest way to run GLM 5.x locally that's not a unified memory system? — https://www.reddit.com/r/LocalLLaMA/comments/1u7y1df/cheapest_way_to_run_glm_5x_locally_thats_not_a/ — Question about cheap local GLM hosting, useful forum topic but not memo material.
- [R] [reddit-localllama] Zhipu surges 33% as Wall Street raises bets on China AI after Anthropic curbs — https://www.reddit.com/r/LocalLLaMA/comments/1u7ozr6/zhipu_surges_33_as_wall_street_raises_bets_on/ — Stock-move headline about Zhipu without direct builder actionability.
- [R] [reddit-localllama] GLM 5.2 API is live, weights are on HF, and ollama has it already — https://www.reddit.com/r/LocalLLaMA/comments/1u7o9vp/glm_52_api_is_live_weights_are_on_hf_and_ollama/ — Useful GLM packaging note, but still duplicate launch coverage.
- [R] [reddit-localllama] Elias in the Lighthouse, Again? Diagnosing Low Diversity in LLM Stories — https://www.reddit.com/r/LocalLLaMA/comments/1u7z9lk/elias_in_the_lighthouse_again_diagnosing_low/ — Interesting paper pointer on story diversity, but not high enough leverage for this memo.
- [R] [reddit-localllama] GLM-5.2 Takes #2 Spot on WebDew Arena — https://www.reddit.com/r/LocalLLaMA/comments/1u7lp1k/glm52_takes_2_spot_on_webdew_arena/ — Yet another GLM leaderboard placement thread, already covered.
- [R] [reddit-localllama] Someone awhile ago did a quant shootout for Qwen3.6, I did shoddy math on it (again) — https://www.reddit.com/r/LocalLLaMA/comments/1u7vdey/someone_awhile_ago_did_a_quant_shootout_for/ — Personal quantization math thread on Qwen, too thin for memo inclusion.
- [R] [reddit-localllama] GLM-5.2 just dropped open weights and it already looks weirdly strong for coding — https://www.reddit.com/r/LocalLLaMA/comments/1u7p25o/glm52_just_dropped_open_weights_and_it_already/ — Early coding impressions of GLM-5.2, but still duplicative launch chatter.
- [R] [reddit-localllama] Looking for a locally-hosted option to create English subtitles (.srt) from video files — https://www.reddit.com/r/LocalLLaMA/comments/1u82cty/looking_for_a_locallyhosted_option_to_create/ — Local subtitle tooling question thread, not a concrete release or report.
- [R] [reddit-localllama] A Year Building a Fully Local Home Voice Assistant · Fulloch — https://www.reddit.com/r/LocalLLaMA/comments/1u84wsg/a_year_building_a_fully_local_home_voice/ — Local voice assistant build retrospective may be interesting, but it is not a top-signal item today.
- [R] [reddit-localllama] Scaling former VibeThinker-1.5B to 3B — now it reaches frontier math & coding performance — https://www.reddit.com/r/LocalLLaMA/comments/1u7dzdr/scaling_former_vibethinker15b_to_3b_now_it/ — Another isolated model-performance post with no broader validation.
- [R] [reddit-localllama] “Wait,” in reasoning models makes my eye twitch — https://www.reddit.com/r/LocalLLaMA/comments/1u7pfu2/wait_in_reasoning_models_makes_my_eye_twitch/ — Reasoning-model style complaint, not a substantive update.
- [R] [reddit-localllama] I didn't know it was possible to compile llamacpp to run cuda + vulkan at the same time.. — https://www.reddit.com/r/LocalLLaMA/comments/1u7ru3d/i_didnt_know_it_was_possible_to_compile_llamacpp/ — Interesting llamacpp build detail, but too anecdotal for the memo.
- [P] [hn-top] GLM-5.2 is the new leading open weights model on Artificial Analysis — https://artificialanalysis.ai/articles/glm-5-2-is-the-new-leading-open-weights-model-on-the-artificial-analysis-intelligence-index — Artificial Analysis puts GLM-5.2 at the top of open weights, with better agentic scores but heavier token use.
- [R] [hn-top] Hacker News but for Independent Blogs — https://bubbles.town/ — Independent-blog HN clone is novel, but it does not change work tomorrow morning.
- [R] [hn-top] U.S. Science Is in Chaos — https://www.scientificamerican.com/article/americas-compact-between-science-and-politics-is-broken/ — Science funding turmoil is important broadly, but outside this memo’s practical scope.
- [R] [hn-top] The founder's playbook: Building an AI-native startup — https://claude.com/blog/the-founders-playbook — Anthropic founder playbook is polished strategy content, but too promotional for this memo.
- [R] [hn-top] Subterranean fungi networks more than 100 quadrillion km in length — https://www.theguardian.com/science/2026/jun/11/arbuscular-mycorrhizal-fungi-plant-life-climate-global-mapping-study — Fungi mapping study is interesting science, not a developer decision item.
- [R] [hn-top] A Nipkow Disk Mechanical TV Simulator — https://analogtv.net/mechanical-lab — Mechanical TV simulator is fun hacker material, but not memo-grade leverage.
- [P] [lobsters] I Could've Rickrolled the Entire FIFA World Cup. All I Needed Was My ID — https://bobdahacker.com/blog/fifa-hack — Researcher says FIFA exposed live stream controls and match systems behind client-side auth checks during the World Cup.
- [P] [lobsters] Making GHC upgrades easy — https://blog.haskell.org/making-ghc-upgrades-easy/ — The GHC team says newer compiler releases should become much easier to adopt by loosening base/GHC coupling.
- [R] [lobsters] Color picking Oklch for mortals — https://hugodaniel.com/posts/color-picking-oklch/ — Useful design explainer on Oklch, but not strong enough against today’s higher-signal items.
- [M] [claude-status] Elevated errors for Claude Opus 4.8 — https://status.claude.com/incidents/k8563n9wd0r8 — Anthropic resolved a Claude Opus 4.8 error spike after affecting chat, API, and coding surfaces.
- [P] [gh-ollama] v0.30.10 — https://github.com/ollama/ollama/releases/tag/v0.30.10 — Ollama 0.30.10 adds MLX support for Command A and North models on Apple Silicon and bumps llama.cpp.
- [R] [gh-crewai] 1.14.8a — https://github.com/crewAIInc/crewAI/releases/tag/1.14.8a — CrewAI 1.14.8a release tag with no visible high-signal change in the queue.