Luke

freedom

[ClaudeDevs @ClaudeDevs]

We've reset 5-hour and weekly rate limits for all users.

[Luke @lukitdev]

Racing to get the complex features done using Fable 5.🤓

[Luke @lukitdev]

So does this mean Fable 5 is very likely not coming to Singapore?

[Luke @lukitdev]

@p2c2kbf Claude’s desktop app is down too.😅

[Luke @lukitdev]

API Error: 500 Internal server error.😅 Claude is down

[Claude @claudeai]

Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors.

Available today at the same price.

[Luke @lukitdev]

@lastz321 体验确实也差些的

[Luke @lukitdev]

@hi_caicai 还挺高频的😂

[Luke @lukitdev]

It looks like access is restricted to certain accounts only.

[Luke @lukitdev]

How to manually enable the 1M context window for Sonnet 4.6 in Claude Code

~/.claude/settings.json:
{
"env": {
"ANTHROPIC_DEFAULT_SONNET_MODEL": "claude-sonnet-4-6[1m]"
}
}

[Luke @lukitdev]

@PanSalut @bridgemindai How to manually enable the 1M context window for Sonnet 4.6 in Claude Code

~/.claude/settings.json:
{
"env": {
"ANTHROPIC_DEFAULT_SONNET_MODEL": "claude-sonnet-4-6[1m]"
}
}

[Luke @lukitdev]

How to manually enable the 1M context window for Sonnet 4.6 in Claude Code

~/.claude/settings.json:
{
"env": {
"ANTHROPIC_DEFAULT_SONNET_MODEL": "claude-sonnet-4-6[1m]"
}
}

[Luke @lukitdev]

@manateelazycat 这是一个用时长换分数的模型

[Luke @lukitdev]

GLM-5 is a total miss for me. These BridgeBench scores are spot on—it’s borderline unusable for actual work. There’s clearly a misalignment in the team’s values if they thought this was ready for prime time.

[BridgeMind @bridgemindai]

BridgeBench is officially live.

The first vibe coding benchmark built by vibe coders, for vibe coders.

Current rankings:

Claude Opus 4.6 — 60.1
MiniMax M2.5 — 59.7
GPT 5.2 Codex — 58.3
Kimi K2.5 — 50.1
GLM 5 — 41.5

Most benchmarks test models in isolation.

BridgeBench tests real coding workflows: debugging, algorithms, refactoring, generation, UI, and security.

We also track what others ignore: cost per task, completion rate, and speed.

MiniMax M2.5 is 0.4 points behind Opus 4.6 at 1/8th the cost.

GLM 5 looks great on paper but only completes 57% of tasks.

Plans to open source soon.

[Luke @lukitdev]

@oran_ge glm 的 coding plan max 套餐也很高频遇到 rate limit error 的报错，对比 kimi 和 minimax 的，可用性差很多

[Luke @lukitdev]

The GLM coding plan is a joke. 🤡 Constant 429 error on a Max subscription. Totally unusable and not worth buying at this stage.

[Luke @lukitdev]

GLM 4.7 has been a letdown. Dealing with constant API instability is a grind. 📉 Fingers crossed for GLM-5.

[Luke @lukitdev]

GLM 4.7 has been a letdown. Dealing with constant API instability is a grind. 📉 Fingers crossed for GLM-5.

[Luke @lukitdev]

Claude Code + Agent Teams is a literal cheat code. 🤯 I’ve become so spoiled by this workflow that every other CLI feels broken now. How did we even work before this?

[Luke @lukitdev]

@Jarodxu7 Happycapy

[Luke @lukitdev]

Time until a productivity explosion 🫠

[Luke @lukitdev]

@Jimmy_JingLv 64G 也是一样不足