Jump to content

AI Sidekick : 6 Fresh Hacks You’ll Actually Use (With Tiny Code)

From JOHNWICK

Practical mini-automations that make your day quieter, smarter, and faster.

1) 60-Second Summaries for Anything Long

Skip the wall of text. This tiny “frequency scorer” picks the most informative sentences — no external APIs.

import re, heapq
from collections import Counter
def summarize(text, max_sents=5):
    sents = re.split(r'(?<=[.!?])\s+', text.strip())
    words = re.findall(r'\w+', text.lower())
    stop = set("""a an the and or but if in on at of to for with by as is are was were be been being from that this it its""".split())
    scores = Counter(w for w in words if w not in stop)
    def sent_score(s): return sum(scores[w] for w in re.findall(r'\w+', s.lower()))
    top = heapq.nlargest(max_sents, sents, key=sent_score)
    return " ".join(top)
print(summarize(open("long_article.txt","r",encoding="utf-8").read(), max_sents=3))

Pro tip: Pipe your meeting transcripts or research PDFs through this and stash the output in Notion.

2) Inbox Calm: Smart “Triage” Labels

A lightweight rules-plus-scoring filter to surface what matters and bury noise.

def triage(subject, body):
    pri = {"invoice":3, "payment":3, "meeting":2, "urgent":5, "resume":2, "offer":4}
    neg = {"sale":-1, "newsletter":-2, "promo":-2, "unsubscribe":-3}
    text = f"{subject} {body}".lower()
    score = sum(v for k,v in pri.items() if k in text) + sum(v for k,v in neg.items() if k in text)
    if score >= 5: return "🔥 Priority"
    if score >= 2: return "✅ Action"
    if score <= -2: return "📨 Low"
    return "📦 Later"
print(triage("Invoice for March", "Please find attached the invoice…"))

Add a dozen keywords from your world (client names, project codes) and watch your inbox behave.

3) Idea Generator, But Structured

Turn “meh” brainstorming into repeatable prompts your model (any LLM) can nail. Framework: P.A.R.A.D.E. - Problem: <1-line pain> - Audience: <who exactly> - Result: <quantified outcome> - Approach: <constraints/tools> - Differentiator: <what’s non-obvious> - Examples: <2 inspiring references> Ask: "Generate 10 ideas strictly following P.A.R.A.D.E., return a table."

You’ll get ideas you can actually test tomorrow morning.

4) Auto-Tag Notes with Real Keywords (TF-IDF Mini)

Great for research dumps, meeting notes, and tweet drafts.

from sklearn.feature_extraction.text import TfidfVectorizer
docs = [open("note1.txt").read(), open("note2.txt").read()]
tfidf = TfidfVectorizer(max_features=2000, ngram_range=(1,2), stop_words="english")
X = tfidf.fit_transform(docs)
def keywords(doc_id, k=8):
    row = X[doc_id].toarray().ravel()
    idx = row.argsort()[-k:][::-1]
    return [tfidf.get_feature_names_out()[i] for i in idx]
print(keywords(0))

Use these tags to name files, create folders, or seed content outlines.

5) “Context Snapshot” Before Any Decision

A simple checklist that prevents dumb mistakes (and it’s shockingly effective):

  • Goal: What is the single outcome?
  • Timebox: What’s the limit (90 minutes, 2 iterations)?
  • Evidence: 3 signals for, 3 against.
  • Risk: What breaks? Cost to revert?
  • Next smallest test: What can you validate in ❤0 minutes?

“The most dangerous decisions are the ones that feel obviously right.” — Naval Ravikant Feed this as context to your AI before asking for a plan — you’ll get sharper, more grounded output.

6) Voice Notes → Action Items (Tiny Offline Pipeline)

Record, transcribe locally (any tool you like), then auto-extract tasks with verbs.

import re
def extract_tasks(transcript):
    verbs = r"(send|email|draft|prepare|schedule|review|call|write|deploy|fix|update)"
    lines = re.split(r'[.\n]', transcript.lower())
    tasks = [l.strip().capitalize() for l in lines if re.search(verbs, l)]
    return [t for t in tasks if len(t) > 8]
sample = "Call Sarah about contract. Draft email to finance. Lunch w/ Ali."
print(extract_tasks(sample))

Dump the result into your task manager via CLI and you’ve got instant momentum.

Wrap-Up

None of these require a PhD or a fat GPU. They’re small, repeatable habits — the kind that stack compound interest in time saved and clarity gained. Start with the summary + triage pair, then layer in tags and voice-to-tasks. Your future self will high-five you.

Read the full article here: https://alitanookhshahid.medium.com/ai-sidekick-6-fresh-hacks-youll-actually-use-with-tiny-code-9a48c8ca743a