project rollup
Default
18996
mentions
across 19 keywords
across 19 keywords
net sentiment
+24%
5882 positive
11691 neutral
1423 negative
volume & sentiment over time
May 13 – Oct 17, 2026 · 4-day buckets
1–8 of 8 keywords
| keyword | mentions | net | trend | top theme | srcs | last seen | |
|---|---|---|---|---|---|---|---|
| timeline | 18695 | +61% | — | 1 | 28m | ||
| Opus 5.5 model documentation | 51 | +61% | claude / pricing | 2 | 2w | ||
| Opus 5.5 official announcement | 50 | +68% | pricing / model | 1 | 2w | ||
| SynthID | 16 | +100% | gemini / google | 1 | 2w | ||
| Cline | 149 | +72% | code | 3 | 2w | ||
| Quill | 32 | -10% | performance | 4 | 2w | ||
| Google SynthID AI image watermark discussion | 1 | +100% | — | 1 | 2mo | ||
| Opus 5.5 versus Claude Opus models | 2 | +100% | — | 1 | 3mo |
mentions
Showing 18926–18950 of 18996 mentions
-
Ask HN: What's the most economical approach to the most tokens?I'm doing web developement, and game development for a hobby project. I've tried lots of harnesses / IDE's - Best I've found is VSCodium.+ Cline + Openrouter, using discounted models (GLM 5.3 Flash is 50% off atm for example) I used Cursor for the last year - it's been decent. But I'm unhappy with the direction their g… I'm doing web developement, and game development for a hobby project. I've tried lots of harnesses / IDE's - Best I've found is VSCodium.+ Cline + Openrouter, using discounted models (GLM 5.3 Flash is 50% off atm for example) I used Cursor for the last year - it's been decent. But I'm unhappy with the direction their going (it's largely just Grok Code at this stage...) Nobody advertises token limits - so it's pretty hard to compare, if you've used multiple tools in the last couple of months - what's your experience been? Will I be able to beat my Openrouter setup on cost per token via a useful model by subscribing to one of these? In my opinion - Grok 4.6 seems comperable with GLM 5.3 Flash. Deepseek 4.1 is my daily driver. I use Sol or Fable for hard problems - but it's expensive man... Please share your opinions.view source ↗
-
Quill's markdown export is broken for nested lists. Reported it twice, no response. Quill's markdown export is broken for nested lists. Reported it twice, no response.view source ↗
-
https://t.co/yo1DHXo2nx https://t.co/yo1DHXo2nxview source ↗
-
Show HN: Slowave – local adaptive memory for coding agentsI started building Slowave because I kept running into the same problem with coding agents: every new session has the codebase and some documentation but not the context behind it, decisions that brought you there and especially the thinking process behind the code. Most memory system solutions focus primarily on the s… I started building Slowave because I kept running into the same problem with coding agents: every new session has the codebase and some documentation but not the context behind it, decisions that brought you there and especially the thinking process behind the code. Most memory system solutions focus primarily on the storage and retrieval aspects (vector search/RAG/graphs/ Markdown files, etc.). After months of storing memories (coding 8+ hours a day produces a lot of memories) these might hallucinate your reasoning model and they can clutter your context window. To treat semantic relationship, such as contradiction, supersession, etc, most systems added an extra LLM layer that summarize memories and continuously evaluate their semantic relevance. That comes with a non-negligible cost and introduces a split-brain system, where a second model is making decisions about memory independently of the agent actually using it. I started looking up into how human brain works, and the first thing striking me was that retrieval is just a part of the whole memory problem. Brain memories are a constant flow of information where what matters gets reinforced, what doesn't decays over time. What really matters for an efficient memory system is to retrieve memories that actually help (a human or an agent) to achieve its current task or goal given the current context. Everything else should be treated as noise. Slowave is my attempt to approach this problem differently: It instructs your coding agent to participate in maintaining its own memory. Each task becomes a feedback loop between your agent and the memory layer: remember -> recall -> use -> feedback -> reinforce / weaken -> decay Your agent tells Slowave whether retrieved memories were useful, irrelevant or stale. Slowave uses that signal to adapt those memories salience. Retrieval works upon this continuous loop of feedback, reinforcement or decay. This means Slowave doesn't need a separate LLM or LLM judge for memory maintenview source ↗
-
Been using Quill for a month. The speed is genuinely impressive, opens instantly even with 5000 notes. Been using Quill for a month. The speed is genuinely impressive, opens instantly even with 5000 notes.view source ↗
-
The pricing jumped to $12/mo which feels steep for a notes app. Otherwise solid. The pricing jumped to $12/mo which feels steep for a notes app. Otherwise solid.view source ↗
-
Show HN: Quill – a fast local-first note appview source ↗
-
So you want to use OpenRouter?If you only care about open source models, cline and opencode provide usage based access and subscriptions for general API access If you only care about open source models, cline and opencode provide usage based access and subscriptions for general API accessview source ↗
-
We have a year to fix security everywhereIt is and it isn't. Why are you comparing the m5max instead of the m4ultra? The big deal to me is the number of compute cores for prefill tps, which is suppose to be 4x faster on the m5ultra. It's my opinion that the m5 ultra is going to be a really big deal in terms of local AI accessibility. Flash sized models (~200-… It is and it isn't. Why are you comparing the m5max instead of the m4ultra? The big deal to me is the number of compute cores for prefill tps, which is suppose to be 4x faster on the m5ultra. It's my opinion that the m5 ultra is going to be a really big deal in terms of local AI accessibility. Flash sized models (~200-300b params) are going to be reasonably fast as long as you aren't throwing 40k context at it on each or the first request (ie, agentic harnesses). Even agentic harnesses like Cline should move at a reasonable clip on m5 ultra. I suppose we will know sooner than later. FYSA: Former m4 ultra 512GB owner and current 4x rtx6000 owner here. I upgraded because I needed more prompt processing speed and concurrency.view source ↗
-
Which tools do Claude, Codex and Cursor choose? We measured 17k runs to find outThey said as of aug 23rd it's 100% rollout. I've noticed 0 improvement. If anything a downgrade. The agents often refuse to use the tools after 1 try because the results are so trash. An open ticket that shows my biggest gripe with the write tool https://github.com/cline/cline/issues/13276 I guess the mistake count is … They said as of aug 23rd it's 100% rollout. I've noticed 0 improvement. If anything a downgrade. The agents often refuse to use the tools after 1 try because the results are so trash. An open ticket that shows my biggest gripe with the write tool https://github.com/cline/cline/issues/13276 I guess the mistake count is down though. But that's because it's bypassing all the tools and just running commandsview source ↗
-
Which tools do Claude, Codex and Cursor choose? We measured 17k runs to find outCline just recently fully upgraded their harness, see here: https://x.com/cline/status/2095897914493243512?s=20 Try if you have a better experience now! Cline just recently fully upgraded their harness, see here: https://x.com/cline/status/2095897914493243512?s=20 Try if you have a better experience now!view source ↗
-
Which tools do Claude, Codex and Cursor choose? We measured 17k runs to find outBecause they're thinking like I did going into the article. Harnesses like Claude expose "tools" to the agent. I usually use Cline but I'm giving up on it for this exact reason. Cline tells the model "you tell me to write a file, I'll get it done" and then it messes everything up, causes tones of errors, and the model … Because they're thinking like I did going into the article. Harnesses like Claude expose "tools" to the agent. I usually use Cline but I'm giving up on it for this exact reason. Cline tells the model "you tell me to write a file, I'll get it done" and then it messes everything up, causes tones of errors, and the model goes "wow that's a broken tool. I'm going to write a python script to write the file instead"view source ↗
-
Ask HN: Why were OpenAI, Claude, and Grok simultaneously down?Also from Cline: After 1177 B.C.: The Survival of Civilizations (2024), ISBN 978-0691192130 Also from Cline: After 1177 B.C.: The Survival of Civilizations (2024), ISBN 978-0691192130view source ↗
-
Ask HN: Why were OpenAI, Claude, and Grok simultaneously down?1177 B.C.: The Year Civilization Collapsed by Eric H. Cline is great and is exactly on this topic. 1177 B.C.: The Year Civilization Collapsed by Eric H. Cline is great and is exactly on this topic.view source ↗
-
Muse Spark 1.3Cline experiment: https://x.com/cline/status/2085237843379519737 Muse code: https://developer.meta.com/ai/resources/blog/build-with-muse... > Co-trained with the harness. Muse Code was in the training loop from day one, so tool calls succeed and plans execute cleanly. Crucially, we trained across multiple harnesses, so… Cline experiment: https://x.com/cline/status/2085237843379519737 Muse code: https://developer.meta.com/ai/resources/blog/build-with-muse... > Co-trained with the harness. Muse Code was in the training loop from day one, so tool calls succeed and plans execute cleanly. Crucially, we trained across multiple harnesses, so while the model is at its best in Muse Code, it still generalizes to other coding agents you already use.view source ↗
-
Muse Spark 1.3You should use their harness. They trained it on multiple harnesses but have specifically optimized it for their harness. Cline also did an independent experiment w spark 1.2 where using the native harness makes it use fewer tokens / turns to accomplish tasks You should use their harness. They trained it on multiple harnesses but have specifically optimized it for their harness. Cline also did an independent experiment w spark 1.2 where using the native harness makes it use fewer tokens / turns to accomplish tasksview source ↗
-
LongCat-2.0 is now free to try in clineview source ↗
-
Ask HN: Who wants to be hired? (September 2026)I’ve been researching how LLMs generate code since summer 2023. My main public artifact is this repository: https://github.com/danvoronov/CodeWithLLM-Updates It’s a living research collection with 400+ commits covering practical experiments, usage patterns, step-by-step guides, model comparisons, AI coding tool evaluat… I’ve been researching how LLMs generate code since summer 2023. My main public artifact is this repository: https://github.com/danvoronov/CodeWithLLM-Updates It’s a living research collection with 400+ commits covering practical experiments, usage patterns, step-by-step guides, model comparisons, AI coding tool evaluations (Cursor, Aider, Cline, Windsurf, etc.). A large part of my work is *exploratory*: navigating large and messy information spaces, testing things hands-on, comparing different approaches, and identifying the small amount of signal that is actually useful. My strongest skill is broad technical *investigation*. I can map an entire problem space rather than focus on a single tool, model, or hypothesis: find the relevant projects, papers, tools, workflows, discussions, benchmarks, and practical examples; test and compare what matters; and organize the results into a clear picture of the field. The goal is not to arrive with a predetermined answer or recommend one particular tool. I can cover the *surrounding landscape* as thoroughly as possible and provide the evidence, comparisons, and structure. Your team can then decide which conclusions matter and what to do with them. I can deliver the results as a structured PDF or HTML research artifact, or discuss the findings in a call. Depending on the project, I can also combine both formats: prepare a written report with the relevant evidence and comparisons, then walk your team through it and answer questions live. I’m interested in projects where someone needs a broad, hands-on investigation of AI-assisted software development: understanding what exists, how different approaches compare, where they work, where they fail, and what is worth paying attention to. I don’t know the conventional job title for this niche. If your team needs someone who can cover a large and changing technical field, turn scattered information into a structured overview, and give you a solid basis for making decisions, I’d be interview source ↗
-
My local model setup on an M4 Pro Mac MiniRecent performance data on my M1 Max 32GB MacBook using oMLX. I have been working on identifying suitable model and config for my use case and system. Using a refactor and suggest improvements prompt for a specific Django code block using VSCode Cline extension. --- Qwen3.8-27B-4bit, Prompt Processing (PP) 66.3 tok/s, … Recent performance data on my M1 Max 32GB MacBook using oMLX. I have been working on identifying suitable model and config for my use case and system. Using a refactor and suggest improvements prompt for a specific Django code block using VSCode Cline extension. --- Qwen3.8-27B-4bit, Prompt Processing (PP) 66.3 tok/s, Token Generation (TG) 11.8 tok/s Ornith-1.5-35B-A3B-MLX-4bit, PP 379.7, TG 45.8 Ornith-1.5-35B-A3B-MLX-4bit, PP 381.5, TG 46.4 Qwen3.6-35B-A3B-mxfp4, PP 389.6, TG 47.6 Qwen3.6-35B-A3B-OptiQ-4bit, PP 342.6, TG 44.4 --- Qwen3.8-27B-4bit generally runs out of output token before completing the task though excellent partial results. Ornith-1.5-35B-A3B-MLX-4bit seems to get in the loop often specially with tool calls. Qwen3.6-35B-A3B-mxfp4 seems to be optimal with speed and quality output. I am going to test Qwen3.6-35B-A3B-4bit soon with same code block just to check my intuition that any derivatives don't seem to perform better than the originals.view source ↗
-
Show HN: Skills MCPJust shipped: skills-mcp One MCP server. ~7,000 AI agent skills. Search → preview → install into Cursor, Claude Code, Claude Desktop, Cline, Windsurf. npx -y @gengirish/skills-mcp Open source, MIT. github.com/gengirish/skills-mcp #BuildWithAIGiri Just shipped: skills-mcp One MCP server. ~7,000 AI agent skills. Search → preview → install into Cursor, Claude Code, Claude Desktop, Cline, Windsurf. npx -y @gengirish/skills-mcp Open source, MIT. github.com/gengirish/skills-mcp #BuildWithAIGiriview source ↗
-
Anomalyco/OpenCode, Aider-AI/aider, and Cline ReviewTried Cline with a local Qwen2.5-Coder 14B model through llama.cpp. Same story. The model works perfectly when called directly, but the agent integration doesn't actually execute the tools. It generates things like: <tools> {"name": "read_files", ...} </tools> as plain text and then just sits there until I tell it to "… Tried Cline with a local Qwen2.5-Coder 14B model through llama.cpp. Same story. The model works perfectly when called directly, but the agent integration doesn't actually execute the tools. It generates things like: <tools> {"name": "read_files", ...} </tools> as plain text and then just sits there until I tell it to "do it" again. It repeats the exact same fake tool call. So instead of: Qwen → tool call → grep/find/read → result → Qwen I get: Qwen → pretend tool call → nothing happens → repeat.view source ↗
-
Anomalyco/OpenCode, Aider-AI/aider, and Cline Reviewview source ↗
-
Literary Fiction Has Lost the PlotFor the record, I haven't read either of the two books discussed here ("Switzy" by Emma Cline and "Ply" by Hernan Diaz). And the one book of literary fiction i did read this month (Nelio Biedermann's much-hyped "Lazar") did not fall into these traps and is wonderful. You can find the first page of that one in Google Pr… For the record, I haven't read either of the two books discussed here ("Switzy" by Emma Cline and "Ply" by Hernan Diaz). And the one book of literary fiction i did read this month (Nelio Biedermann's much-hyped "Lazar") did not fall into these traps and is wonderful. You can find the first page of that one in Google Preview if you're curious.view source ↗
-
⚔ GLM-5.3-Flash Aims Straight at the Post-Hike DeepSeek Zhipu's GLM-5.3-Flash — revealed this week as the anonymous "Ox Alpha" — has been open-weighted and priced at roughly one-tenth of GLM-5.3. Much of the early discussion compares it to DeepSeek's V4 Flash, which recently raised prices. Zhihu contributor 起步十档, who … ⚔ GLM-5.3-Flash Aims Straight at the Post-Hike DeepSeek Zhipu's GLM-5.3-Flash — revealed this week as the anonymous "Ox Alpha" — has been open-weighted and priced at roughly one-tenth of GLM-5.3. Much of the early discussion compares it to DeepSeek's V4 Flash, which recently raised prices. Zhihu contributor 起步十档, who ran Ox Alpha inside real workflows before the reveal, gives a practitioner's verdict in one line: it is built to kill the post-hike DeepSeek. His case rests on three legs — performance, token efficiency, and an architecture change that makes the price possible. 1️⃣ It clears the bar for long-horizon work Official scores put Flash between Grok 4.6 and GLM-5.3, and clearly ahead of DeepSeek V4 Flash. In the author's own testing, its frontend ability roughly matches an early gray-test build of DeepSeek V4 Pro, while its backend is noticeably weaker than GLM-5.3 — but still usable on long-horizon tasks as long as the connection holds. His rule of thumb: any model past the Claude Opus 4.6 line is workflow-ready for long tasks. Beyond that, differences come down to reasoning style and accuracy, not viability. One caveat he flags: during the anonymous test the deployment was unstable, and some believe it served a mid-training checkpoint rather than the final model. 2️⃣ The real weapon: token efficiency Comparing peak API prices against the post-hike DeepSeek V4 Flash, the author notes cached input is actually 2x more expensive, while regular input and output sit at roughly 26% of DeepSeek's price. Since cached input is often the bulk of the bill, he wants real-world tests before calling a winner on price alone. But his own usage points the same direction. In one to two hours of real work — reading and writing files, running tests — Ox Alpha burned barely over 100K tokens. He estimates DeepSeek would need 250-300K for the same workload. His prediction: same tasks, run on both APIs, will come out cheaper on Flash — with clearly better performance. 3️⃣ His unview source ↗
-
Ask HN: Are you still using AI code autocomplete?With all the recent progress in coding assistants, I've been wondering whether AI-powered autocomplete is still an important part of developers' workflows. I personally never got into the habit of using it very heavily. By the time I started having access to decent AI autocomplete tools, things like Cline and Augment w… With all the recent progress in coding assistants, I've been wondering whether AI-powered autocomplete is still an important part of developers' workflows. I personally never got into the habit of using it very heavily. By the time I started having access to decent AI autocomplete tools, things like Cline and Augment were already around, so I ended up relying more on assistants instead. That made me curious about how other people use it today. Do you still use Cursor Tab, Copilot, or something similar? If you also use coding assistants, do you find yourself using AI autocomplete less, or do the two complement each other?view source ↗