Claude Sonnet 5.5 — drop-everything. Anthropic's September 28 announcement says the model outputs 30%+ faster and costs up to 30% less per task, while per-token pricing stays unchanged. A direct Hacker News thread had 851 points and 582 comments at snapshot. Builder significance: compare accepted outcomes, tokens, wall time and corrections on a fixed task. The benchmark and savings figures are vendor reported.
Two concise reads that explain why this digest matters.
Dots. OpenAI's primary describes always-on agents and distinguishes read-only proactive research from tasks governed by permissions and review. Direct HN had 444 points / 337 comments after a 10:07 PDT submission; specialist Reddit showed 727 indexed votes. Builder significance: inspect the read/write boundary and retained-memory controls. The Help Center describes gradual rollout and optional local access. No independent permissions test was performed.
SourceGPT-6.1 Sol. Official announcement lists standard API prices of $2 input / $0.10 cached input / $10 output per million tokens and vendor-reported near-Astra evaluations. HN had 747 / 685 after a 10:06 PDT submission. Builder significance: compare accepted-task cost, retries and correction time under the same harness. Available in Work/Codex/API; ordinary Chat availability was excluded by the primary at refresh. No independent savings comparison was performed.
SourceClaude Sonnet 5.5 — drop-everything. Anthropic's September 28 announcement says the model outputs 30%+ faster and costs up to 30% less per task, while per-token pricing stays unchanged. A direct Hacker News thread had 851 points and 582 comments at snapshot. Builder significance: compare accepted outcomes, tokens, wall time and corrections on a fixed task. The benchmark and savings figures are vendor reported.
SourceOpenAI Pro allowance change — single segment. An OpenAI executive's announcement had 11.55 million views and 6,645 replies; follow-up clarification described new 5× and 10× tiers and temporary treatment for existing 20× subscribers. Builder significance: plan limits affect real agent workflow capacity. Official Help Center copy had not yet reconciled with the posts at snapshot; verify current account terms before deciding.
SourceWorld Labs joining AMD — single segment. World Labs' primary account describes a definitive agreement expected to close by year-end, while AMD's release reports an approximately $8.2 billion all-stock deal. HN discussion had 295 points and 111 comments. Builder significance: watch whether spatial-model workloads shape measurable compute-system designs; the deal itself proves no performance gain.
SourceHigh-discussion omission — Jeff. Jeff's repository drew 546 HN points and 212 comments, but its 28 ms decision latency is author reported and no independent accuracy test was found. Kept at mention until replicated.
SourceOther omissions. NVIDIA's agent-safety platform has useful permission architecture but no independent containment test here. Meta Muse for Small Business is new, but comparable early discussion metrics were missing. A September 16 advertiser-leakage paper.pdf) had fresh HN discussion but is not a new launch.
SourcePosts worth preserving because they point at repeatable builder behavior.
Open-source proof objects with reusable product or workflow patterns.
Commercial surfaces that show where attention or buyer pain is collecting.
The short action layer from the feed.
Direct public sources used by this digest.