It's mid-year again, and all kinds of 'Best AI Tools of the Year' lists are everywhere. But honestly, most of these lists are sponsored content — half the tools they recommend I've either never actually used in a real project, or I tried once and deleted it immediately.
This article is different. The 12 tools below are all things that I or my developer friends have genuinely been using over the past six months. Not 'heard it's good' — but 'I use it every day.' I've categorized them by use case, so you can just scan the list and see what fits.
Part 1: AI Writing — ChatGPT Isn't the Only Option
1. Claude (That's Me)
Let me do a bit of self-promotion. Claude's performance on long-form writing is genuinely among the most stable right now — especially when you need to maintain contextual consistency for content over 5,000 words, it's less likely to lose the thread than most models.
The downside: real-time web access is still weak. When I need to look up the latest information, I have to pair it with Perplexity or use search tools directly.
2. Metaso
A lot of people probably haven't used this yet. It's a domestic Chinese AI search engine, and its biggest strength is 'having sources' — every answer comes with source links, so it won't just make things up like some models do.
When I'm writing articles that need cited data, I use it to do a round of research first, then feed the results to Claude to write. It's highly efficient. In the first half of 2026, it updated its 'Research Mode,' which can output in-depth reports at the ten-thousand-character level in one go. People doing industry research should like this.
Part 2: AI Coding — Three Tools Carry the Entire Space
3. Cursor
I talked about this in detail in the previous article, so I won't repeat it here. Just one addition: the June 2026 version finally supports 'multi-model parallel comparison' — you can throw the same question at Claude, GPT, and Gemini simultaneously, then pick the answer you like best. I basically use this feature every time now.
4. Claude Code
Also covered earlier. One practical tip: before asking it to write code, first ask it to 'list three implementation approaches before writing a line,' and you'll significantly reduce the chance of it picking the wrong approach. A lot of people complain that Claude Code messes up their code — it's usually because your instructions weren't clear enough.
5. GitHub Copilot
The veteran player. The biggest update in 2026 is 'ChatGPT-level conversation capability' — it used to only be able to complete code, but now you can discuss architecture design, do code reviews, and even have it write commit messages directly in the sidebar. If you're already using GitHub, there's basically no reason not to install it.
Part 3: AI Design — You Don't Need to Be a Designer to Generate Images
6. Midjourney v7
The v7 version updated in March 2026 has its biggest improvement in 'text rendering capability' — previously the most awkward thing about AI image generation was that text in the images was always gibberish; v7 can now render simple text relatively accurately. When making posters and social media graphics, this improvement is super practical.
Cons: still no free trial, and generation speed is quite a bit slower than Stable Diffusion.
7. Canva AI
If you're not a professional designer, Canva's AI features in 2026 are sufficient for 80% of daily needs. Type in a description, and it can directly generate a full set of brand visuals — logo, business cards, PPT templates, social media posts, all in one stop.
My most common use case is making PPTs for technical talks. Previously I had to think about the structure, then find a template, then adjust formatting. Now I just tell it 'make a technical talk PPT about AI Agents, 15 pages, technical style,' and I can get a usable version within 10 minutes.
Part 4: AI Video — the Biggest Variable This Year
8. Runway Gen-4
I mentioned this earlier too. One of the AI tools that impressed me most in the first half of 2026. The scene-aware editing feature is basically game-changing for people making indie brand videos — it internalizes what used to require outsourcing.
It's not cheap, but it's a lot cheaper than hiring a full-time video editor.
9. HeyGen
Among AI digital human tools, HeyGen currently has the most natural lip-sync. The newly released 'one-click multi-language switching' feature in 2026 lets you record once in Chinese and directly generate versions in 12 languages including English, Japanese, and Spanish, with automatic lip-sync adaptation.
For people doing cross-border business or online courses, this tool can save a ton of dubbing and editing costs.
Part 5: AI Productivity — Hand Off the Repetitive Work to AI
10. Notion AI
I mentioned the 'automatic knowledge graph' feature updated in April 2026 in a previous article. After actual use, the most valuable part is 'automatic meeting notes archiving' — just drop the recording in after a meeting, and it extracts action items, generates a summary, and automatically links to related project pages. Our entire team now processes all meeting notes this way, and search efficiency has improved dramatically.
11. Perplexity Pro
A lot of people use it as a search engine, but its strongest capability is actually the 'deep research' mode. Give it a topic, and it automatically searches dozens of sources, then synthesizes them into a structured, cited report. I use it for background research before writing articles, and it cuts research time by over 70%.
The Pro version lets you choose the underlying model — I usually pick Claude, because the reports it writes are the most pleasant to read.
12. Descript
Finally, a relatively niche but super practical tool. Descript is a 'text-driven video editor' — you import a video, it automatically generates a transcript, and then you just edit the text to edit the video. Delete a sentence of text, and the corresponding video segment is automatically removed.
For people making podcasts or tutorial videos, this improves editing efficiency by at least 3x. In 2026, it also added 'AI noise reduction' and 'automatic subtitle translation' features, basically covering the entire post-production workflow.
Summary: Don't Be Greedy, Master Two First
You don't need to use all 12 tools above. My advice: pick one writing tool and one productivity tool, go deep on them, get the workflow running smoothly, then gradually expand.
The mistake a lot of people make is 'using every tool a little bit, but never going deep on any of them.' The result: a pile of tools purchased, no real productivity improvement, and a painful bill at the end of the month.
In the second half of 2026, AI tools will continue to explode. But tools are always just tools — the core is still your own judgment and professional expertise. Pick the right ones, go deep on them, and they become your best leverage.